linux.git
14 months agoKVM: Drop unused @may_block param from gfn_to_pfn_cache_invalidate_start()
Sean Christopherson [Tue, 5 Mar 2024 00:37:42 +0000 (16:37 -0800)]
KVM: Drop unused @may_block param from gfn_to_pfn_cache_invalidate_start()

Remove gfn_to_pfn_cache_invalidate_start()'s unused @may_block parameter,
which was leftover from KVM's abandoned (for now) attempt to support guest
usage of gfn_to_pfn caches.

Fixes: a4bff3df5147 ("KVM: pfncache: remove KVM_GUEST_USES_PFN usage")
Reported-by: Like Xu <like.xu.linux@gmail.com>
Cc: Paul Durrant <paul@xen.org>
Cc: David Woodhouse <dwmw2@infradead.org>
Reviewed-by: Paul Durrant <paul@xen.org>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Link: https://lore.kernel.org/r/20240305003742.245767-1-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: selftests: Add coverage of EPT-disabled to vmx_dirty_log_test
David Matlack [Fri, 15 Mar 2024 23:05:41 +0000 (16:05 -0700)]
KVM: selftests: Add coverage of EPT-disabled to vmx_dirty_log_test

Extend vmx_dirty_log_test to include accesses made by L2 when EPT is
disabled.

This commit adds explicit coverage of a bug caught by syzkaller, where
the TDP MMU would clear D-bits instead of write-protecting SPTEs being
used to map an L2, which only happens when L1 does not enable EPT,
causing writes made by L2 to not be reflected in the dirty log when PML
is enabled:

  $ ./vmx_dirty_log_test
  Nested EPT: disabled
  ==== Test Assertion Failure ====
    x86_64/vmx_dirty_log_test.c:151: test_bit(0, bmap)
    pid=72052 tid=72052 errno=4 - Interrupted system call
    (stack trace empty)
    Page 0 incorrectly reported clean

Opportunistically replace the volatile casts with {READ,WRITE}_ONCE().

Link: https://lore.kernel.org/kvm/000000000000c6526f06137f18cc@google.com/
Signed-off-by: David Matlack <dmatlack@google.com>
Link: https://lore.kernel.org/r/20240315230541.1635322-5-dmatlack@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: x86/mmu: Fix and clarify comments about clearing D-bit vs. write-protecting
David Matlack [Fri, 15 Mar 2024 23:05:40 +0000 (16:05 -0700)]
KVM: x86/mmu: Fix and clarify comments about clearing D-bit vs. write-protecting

Drop the "If AD bits are enabled/disabled" verbiage from the comments
above kvm_tdp_mmu_clear_dirty_{slot,pt_masked}() since TDP MMU SPTEs may
need to be write-protected even when A/D bits are enabled. i.e. These
comments aren't technically correct.

No functional change intended.

Signed-off-by: David Matlack <dmatlack@google.com>
Link: https://lore.kernel.org/r/20240315230541.1635322-4-dmatlack@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: x86/mmu: Remove function comments above clear_dirty_{gfn_range,pt_masked}()
David Matlack [Fri, 15 Mar 2024 23:05:39 +0000 (16:05 -0700)]
KVM: x86/mmu: Remove function comments above clear_dirty_{gfn_range,pt_masked}()

Drop the comments above clear_dirty_gfn_range() and
clear_dirty_pt_masked(), since each is word-for-word identical to the
comment above their parent function.

Leave the comment on the parent functions since they are APIs called by
the KVM/x86 MMU.

No functional change intended.

Signed-off-by: David Matlack <dmatlack@google.com>
Link: https://lore.kernel.org/r/20240315230541.1635322-3-dmatlack@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: x86/mmu: Write-protect L2 SPTEs in TDP MMU when clearing dirty status
David Matlack [Fri, 15 Mar 2024 23:05:38 +0000 (16:05 -0700)]
KVM: x86/mmu: Write-protect L2 SPTEs in TDP MMU when clearing dirty status

Check kvm_mmu_page_ad_need_write_protect() when deciding whether to
write-protect or clear D-bits on TDP MMU SPTEs, so that the TDP MMU
accounts for any role-specific reasons for disabling D-bit dirty logging.

Specifically, TDP MMU SPTEs must be write-protected when the TDP MMU is
being used to run an L2 (i.e. L1 has disabled EPT) and PML is enabled.
KVM always disables PML when running L2, even when L1 and L2 GPAs are in
the some domain, so failing to write-protect TDP MMU SPTEs will cause
writes made by L2 to not be reflected in the dirty log.

Reported-by: syzbot+900d58a45dcaab9e4821@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=900d58a45dcaab9e4821
Fixes: 5982a5392663 ("KVM: x86/mmu: Use kvm_ad_enabled() to determine if TDP MMU SPTEs need wrprot")
Cc: stable@vger.kernel.org
Cc: Vipin Sharma <vipinsh@google.com>
Cc: Sean Christopherson <seanjc@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Link: https://lore.kernel.org/r/20240315230541.1635322-2-dmatlack@google.com
[sean: massage shortlog and changelog, tweak ternary op formatting]
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: x86/mmu: Precisely invalidate MMU root_role during CPUID update
Sean Christopherson [Mon, 8 Apr 2024 23:11:15 +0000 (16:11 -0700)]
KVM: x86/mmu: Precisely invalidate MMU root_role during CPUID update

Set kvm_mmu_page_role.invalid to mark the various MMU root_roles invalid
during CPUID update in order to force a refresh, instead of zeroing out
the entire role.  This fixes a bug where kvm_mmu_free_roots() incorrectly
thinks a root is indirect, i.e. not a TDP MMU, due to "direct" being
zeroed, which in turn causes KVM to take mmu_lock for write instead of
read.

Note, paving over the entire role was largely unintentional, commit
7a458f0e1ba1 ("KVM: x86/mmu: remove extended bits from mmu_role, rename
field") simply missed that "invalid" could be set.

Fixes: 576a15de8d29 ("KVM: x86/mmu: Free TDP MMU roots while holding mmy_lock for read")
Reported-by: syzbot+dc308fcfcd53f987de73@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/0000000000009b38080614c49bdb@google.com
Cc: Phi Nguyen <phind.uet@gmail.com>
Link: https://lore.kernel.org/r/20240408231115.1387279-1-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: VMX: Disable LBR virtualization if the CPU doesn't support LBR callstacks
Sean Christopherson [Thu, 7 Mar 2024 01:13:44 +0000 (17:13 -0800)]
KVM: VMX: Disable LBR virtualization if the CPU doesn't support LBR callstacks

Disable LBR virtualization if the CPU doesn't support callstacks, which
were introduced in HSW (see commit e9d7f7cd97c4 ("perf/x86/intel: Add
basic Haswell LBR call stack support"), as KVM unconditionally configures
the perf LBR event with PERF_SAMPLE_BRANCH_CALL_STACK, i.e. LBR
virtualization always fails on pre-HSW CPUs.

Simply disable LBR support on such CPUs, as it has never worked, i.e.
there is no risk of breaking an existing setup, and figuring out a way
to performantly context switch LBRs on old CPUs is not worth the effort.

Fixes: be635e34c284 ("KVM: vmx/pmu: Expose LBR_FMT in the MSR_IA32_PERF_CAPABILITIES")
Cc: Mingwei Zhang <mizhang@google.com>
Cc: Jim Mattson <jmattson@google.com>
Tested-by: Mingwei Zhang <mizhang@google.com>
Link: https://lore.kernel.org/r/20240307011344.835640-4-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoperf/x86/intel: Expose existence of callback support to KVM
Sean Christopherson [Thu, 7 Mar 2024 01:13:43 +0000 (17:13 -0800)]
perf/x86/intel: Expose existence of callback support to KVM

Add a "has_callstack" field to the x86_pmu_lbr structure used to pass
information to KVM, and set it accordingly in x86_perf_get_lbr().  KVM
will use has_callstack to avoid trying to create perf LBR events with
PERF_SAMPLE_BRANCH_CALL_STACK on CPUs that don't support callstacks.

Reviewed-by: Mingwei Zhang <mizhang@google.com>
Link: https://lore.kernel.org/r/20240307011344.835640-3-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: VMX: Snapshot LBR capabilities during module initialization
Sean Christopherson [Thu, 7 Mar 2024 01:13:42 +0000 (17:13 -0800)]
KVM: VMX: Snapshot LBR capabilities during module initialization

Snapshot VMX's LBR capabilities once during module initialization instead
of calling into perf every time a vCPU reconfigures its vPMU.  This will
allow massaging the LBR capabilities, e.g. if the CPU doesn't support
callstacks, without having to remember to update multiple locations.

Opportunistically tag vmx_get_perf_capabilities() with __init, as it's
only called from vmx_set_cpu_caps().

Reviewed-by: Mingwei Zhang <mizhang@google.com>
Link: https://lore.kernel.org/r/20240307011344.835640-2-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: VMX: Ignore MKTME KeyID bits when intercepting #PF for allow_smaller_maxphyaddr
Tao Su [Tue, 19 Mar 2024 03:11:11 +0000 (11:11 +0800)]
KVM: VMX: Ignore MKTME KeyID bits when intercepting #PF for allow_smaller_maxphyaddr

Use the raw/true host.MAXPHYADDR when deciding whether or not KVM must
intercept #PFs when allow_smaller_maxphyaddr is enabled, as any adjustments
the kernel makes to boot_cpu_data.x86_phys_bits to account for MKTME KeyID
bits do not apply to the guest physical address space.  I.e. the KeyID are
off-limits for host physical addresses, but are not reserved for GPAs as
far as hardware is concerned.

Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Link: https://lore.kernel.org/r/20240319031111.495006-1-tao1.su@linux.intel.com
[sean: massage changelog]
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: selftests: fix supported_flags for riscv
Andrew Jones [Wed, 3 Apr 2024 12:33:01 +0000 (14:33 +0200)]
KVM: selftests: fix supported_flags for riscv

commit 849c1816436f ("KVM: selftests: fix supported_flags for aarch64")
fixed the set-memory-region test for aarch64 by declaring the read-only
flag is supported. riscv also supports the read-only flag. Fix it too.

Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
Link: https://lore.kernel.org/r/20240403123300.63923-2-ajones@ventanamicro.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: selftests: fix max_guest_memory_test with more that 256 vCPUs
Maxim Levitsky [Fri, 15 Mar 2024 14:35:07 +0000 (10:35 -0400)]
KVM: selftests: fix max_guest_memory_test with more that 256 vCPUs

max_guest_memory_test uses ucalls to sync with the host, but
it also resets the guest RIP back to its initial value in between
tests stages.

This makes the guest never reach the code which frees the ucall struct
and since a fixed pool of 512 ucall structs is used, the test starts
to fail when more that 256 vCPUs are used.

Fix that by replacing the manual register reset with a loop in
the guest code.

Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Link: https://lore.kernel.org/r/20240315143507.102629-1-mlevitsk@redhat.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: selftests: Verify post-RESET value of PERF_GLOBAL_CTRL in PMCs test
Sean Christopherson [Sat, 9 Mar 2024 01:36:41 +0000 (17:36 -0800)]
KVM: selftests: Verify post-RESET value of PERF_GLOBAL_CTRL in PMCs test

Add a guest assert in the PMU counters test to verify that KVM stuffs
the vCPU's post-RESET value to globally enable all general purpose
counters.  Per Intel's SDM,

  IA32_PERF_GLOBAL_CTRL:  Sets bits n-1:0 and clears the upper bits.

and

  Where "n" is the number of general-purpose counters available in
  the processor.

For the edge case where there are zero GP counters, follow the spirit
of the architecture, not the SDM's literal wording, which doesn't account
for this possibility and would require the CPU to set _all_ bits in
PERF_GLOBAL_CTRL.

Reviewed-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
Tested-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
Link: https://lore.kernel.org/r/20240309013641.1413400-3-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: x86/pmu: Set enable bits for GP counters in PERF_GLOBAL_CTRL at "RESET"
Sean Christopherson [Sat, 9 Mar 2024 01:36:40 +0000 (17:36 -0800)]
KVM: x86/pmu: Set enable bits for GP counters in PERF_GLOBAL_CTRL at "RESET"

Set the enable bits for general purpose counters in IA32_PERF_GLOBAL_CTRL
when refreshing the PMU to emulate the MSR's architecturally defined
post-RESET behavior.  Per Intel's SDM:

  IA32_PERF_GLOBAL_CTRL:  Sets bits n-1:0 and clears the upper bits.

and

  Where "n" is the number of general-purpose counters available in the processor.

AMD also documents this behavior for PerfMonV2 CPUs in one of AMD's many
PPRs.

Do not set any PERF_GLOBAL_CTRL bits if there are no general purpose
counters, although a literal reading of the SDM would require the CPU to
set either bits 63:0 or 31:0.  The intent of the behavior is to globally
enable all GP counters; honor the intent, if not the letter of the law.

Leaving PERF_GLOBAL_CTRL '0' effectively breaks PMU usage in guests that
haven't been updated to work with PMUs that support PERF_GLOBAL_CTRL.
This bug was recently exposed when KVM added supported for AMD's
PerfMonV2, i.e. when KVM started exposing a vPMU with PERF_GLOBAL_CTRL to
guest software that only knew how to program v1 PMUs (that don't support
PERF_GLOBAL_CTRL).

Failure to emulate the post-RESET behavior results in such guests
unknowingly leaving all general purpose counters globally disabled (the
entire reason the post-RESET value sets the GP counter enable bits is to
maintain backwards compatibility).

The bug has likely gone unnoticed because PERF_GLOBAL_CTRL has been
supported on Intel CPUs for as long as KVM has existed, i.e. hardly anyone
is running guest software that isn't aware of PERF_GLOBAL_CTRL on Intel
PMUs.  And because up until v6.0, KVM _did_ emulate the behavior for Intel
CPUs, although the old behavior was likely dumb luck.

Because (a) that old code was also broken in its own way (the history of
this code is a comedy of errors), and (b) PERF_GLOBAL_CTRL was documented
as having a value of '0' post-RESET in all SDMs before March 2023.

Initial vPMU support in commit f5132b01386b ("KVM: Expose a version 2
architectural PMU to a guests") *almost* got it right (again likely by
dumb luck), but for some reason only set the bits if the guest PMU was
advertised as v1:

        if (pmu->version == 1) {
                pmu->global_ctrl = (1 << pmu->nr_arch_gp_counters) - 1;
                return;
        }

Commit f19a0c2c2e6a ("KVM: PMU emulation: GLOBAL_CTRL MSR should be
enabled on reset") then tried to remedy that goof, presumably because
guest PMUs were leaving PERF_GLOBAL_CTRL '0', i.e. weren't enabling
counters.

        pmu->global_ctrl = ((1 << pmu->nr_arch_gp_counters) - 1) |
                (((1ull << pmu->nr_arch_fixed_counters) - 1) << X86_PMC_IDX_FIXED);
        pmu->global_ctrl_mask = ~pmu->global_ctrl;

That was KVM's behavior up until commit c49467a45fe0 ("KVM: x86/pmu:
Don't overwrite the pmu->global_ctrl when refreshing") removed
*everything*.  However, it did so based on the behavior defined by the
SDM , which at the time stated that "Global Perf Counter Controls" is
'0' at Power-Up and RESET.

But then the March 2023 SDM (325462-079US), stealthily changed its
"IA-32 and Intel 64 Processor States Following Power-up, Reset, or INIT"
table to say:

  IA32_PERF_GLOBAL_CTRL: Sets bits n-1:0 and clears the upper bits.

Note, kvm_pmu_refresh() can be invoked multiple times, i.e. it's not a
"pure" RESET flow.  But it can only be called prior to the first KVM_RUN,
i.e. the guest will only ever observe the final value.

Note #2, KVM has always cleared global_ctrl during refresh (see commit
f5132b01386b ("KVM: Expose a version 2 architectural PMU to a guests")),
i.e. there is no danger of breaking existing setups by clobbering a value
set by userspace.

Reported-by: Babu Moger <babu.moger@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Like Xu <like.xu.linux@gmail.com>
Cc: Mingwei Zhang <mizhang@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: stable@vger.kernel.org
Reviewed-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
Tested-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
Link: https://lore.kernel.org/r/20240309013641.1413400-2-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: x86/mmu: x86: Don't overflow lpage_info when checking attributes
Rick Edgecombe [Thu, 14 Mar 2024 21:29:02 +0000 (14:29 -0700)]
KVM: x86/mmu: x86: Don't overflow lpage_info when checking attributes

Fix KVM_SET_MEMORY_ATTRIBUTES to not overflow lpage_info array and trigger
KASAN splat, as seen in the private_mem_conversions_test selftest.

When memory attributes are set on a GFN range, that range will have
specific properties applied to the TDP. A huge page cannot be used when
the attributes are inconsistent, so they are disabled for those the
specific huge pages. For internal KVM reasons, huge pages are also not
allowed to span adjacent memslots regardless of whether the backing memory
could be mapped as huge.

What GFNs support which huge page sizes is tracked by an array of arrays
'lpage_info' on the memslot, of ‘kvm_lpage_info’ structs. Each index of
lpage_info contains a vmalloc allocated array of these for a specific
supported page size. The kvm_lpage_info denotes whether a specific huge
page (GFN and page size) on the memslot is supported. These arrays include
indices for unaligned head and tail huge pages.

Preventing huge pages from spanning adjacent memslot is covered by
incrementing the count in head and tail kvm_lpage_info when the memslot is
allocated, but disallowing huge pages for memory that has mixed attributes
has to be done in a more complicated way. During the
KVM_SET_MEMORY_ATTRIBUTES ioctl KVM updates lpage_info for each memslot in
the range that has mismatched attributes. KVM does this a memslot at a
time, and marks a special bit, KVM_LPAGE_MIXED_FLAG, in the kvm_lpage_info
for any huge page. This bit is essentially a permanently elevated count.
So huge pages will not be mapped for the GFN at that page size if the
count is elevated in either case: a huge head or tail page unaligned to
the memslot or if KVM_LPAGE_MIXED_FLAG is set because it has mixed
attributes.

To determine whether a huge page has consistent attributes, the
KVM_SET_MEMORY_ATTRIBUTES operation checks an xarray to make sure it
consistently has the incoming attribute. Since level - 1 huge pages are
aligned to level huge pages, it employs an optimization. As long as the
level - 1 huge pages are checked first, it can just check these and assume
that if each level - 1 huge page contained within the level sized huge
page is not mixed, then the level size huge page is not mixed. This
optimization happens in the helper hugepage_has_attrs().

Unfortunately, although the kvm_lpage_info array representing page size
'level' will contain an entry for an unaligned tail page of size level,
the array for level - 1  will not contain an entry for each GFN at page
size level. The level - 1 array will only contain an index for any
unaligned region covered by level - 1 huge page size, which can be a
smaller region. So this causes the optimization to overflow the level - 1
kvm_lpage_info and perform a vmalloc out of bounds read.

In some cases of head and tail pages where an overflow could happen,
callers skip the operation completely as KVM_LPAGE_MIXED_FLAG is not
required to prevent huge pages as discussed earlier. But for memslots that
are smaller than the 1GB page size, it does call hugepage_has_attrs(). In
this case the huge page is both the head and tail page. The issue can be
observed simply by compiling the kernel with CONFIG_KASAN_VMALLOC and
running the selftest “private_mem_conversions_test”, which produces the
output like the following:

BUG: KASAN: vmalloc-out-of-bounds in hugepage_has_attrs+0x7e/0x110
Read of size 4 at addr ffffc900000a3008 by task private_mem_con/169
Call Trace:
  dump_stack_lvl
  print_report
  ? __virt_addr_valid
  ? hugepage_has_attrs
  ? hugepage_has_attrs
  kasan_report
  ? hugepage_has_attrs
  hugepage_has_attrs
  kvm_arch_post_set_memory_attributes
  kvm_vm_ioctl

It is a little ambiguous whether the unaligned head page (in the bug case
also the tail page) should be expected to have KVM_LPAGE_MIXED_FLAG set.
It is not functionally required, as the unaligned head/tail pages will
already have their kvm_lpage_info count incremented. The comments imply
not setting it on unaligned head pages is intentional, so fix the callers
to skip trying to set KVM_LPAGE_MIXED_FLAG in this case, and in doing so
not call hugepage_has_attrs().

Cc: stable@vger.kernel.org
Fixes: 90b4fe17981e ("KVM: x86: Disallow hugepages when memory attributes are mixed")
Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
Reviewed-by: Kai Huang <kai.huang@intel.com>
Reviewed-by: Chao Peng <chao.p.peng@linux.intel.com>
Link: https://lore.kernel.org/r/20240314212902.2762507-1-rick.p.edgecombe@intel.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: x86/pmu: Disable support for adaptive PEBS
Sean Christopherson [Thu, 7 Mar 2024 00:58:33 +0000 (16:58 -0800)]
KVM: x86/pmu: Disable support for adaptive PEBS

Drop support for virtualizing adaptive PEBS, as KVM's implementation is
architecturally broken without an obvious/easy path forward, and because
exposing adaptive PEBS can leak host LBRs to the guest, i.e. can leak
host kernel addresses to the guest.

Bug #1 is that KVM doesn't account for the upper 32 bits of
IA32_FIXED_CTR_CTRL when (re)programming fixed counters, e.g
fixed_ctrl_field() drops the upper bits, reprogram_fixed_counters()
stores local variables as u8s and truncates the upper bits too, etc.

Bug #2 is that, because KVM _always_ sets precise_ip to a non-zero value
for PEBS events, perf will _always_ generate an adaptive record, even if
the guest requested a basic record.  Note, KVM will also enable adaptive
PEBS in individual *counter*, even if adaptive PEBS isn't exposed to the
guest, but this is benign as MSR_PEBS_DATA_CFG is guaranteed to be zero,
i.e. the guest will only ever see Basic records.

Bug #3 is in perf.  intel_pmu_disable_fixed() doesn't clear the upper
bits either, i.e. leaves ICL_FIXED_0_ADAPTIVE set, and
intel_pmu_enable_fixed() effectively doesn't clear ICL_FIXED_0_ADAPTIVE
either.  I.e. perf _always_ enables ADAPTIVE counters, regardless of what
KVM requests.

Bug #4 is that adaptive PEBS *might* effectively bypass event filters set
by the host, as "Updated Memory Access Info Group" records information
that might be disallowed by userspace via KVM_SET_PMU_EVENT_FILTER.

Bug #5 is that KVM doesn't ensure LBR MSRs hold guest values (or at least
zeros) when entering a vCPU with adaptive PEBS, which allows the guest
to read host LBRs, i.e. host RIPs/addresses, by enabling "LBR Entries"
records.

Disable adaptive PEBS support as an immediate fix due to the severity of
the LBR leak in particular, and because fixing all of the bugs will be
non-trivial, e.g. not suitable for backporting to stable kernels.

Note!  This will break live migration, but trying to make KVM play nice
with live migration would be quite complicated, wouldn't be guaranteed to
work (i.e. KVM might still kill/confuse the guest), and it's not clear
that there are any publicly available VMMs that support adaptive PEBS,
let alone live migrate VMs that support adaptive PEBS, e.g. QEMU doesn't
support PEBS in any capacity.

Link: https://lore.kernel.org/all/20240306230153.786365-1-seanjc@google.com
Link: https://lore.kernel.org/all/ZeepGjHCeSfadANM@google.com
Fixes: c59a1f106f5c ("KVM: x86/pmu: Add IA32_PEBS_ENABLE MSR emulation for extended PEBS")
Cc: stable@vger.kernel.org
Cc: Like Xu <like.xu.linux@gmail.com>
Cc: Mingwei Zhang <mizhang@google.com>
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Cc: Zhang Xiong <xiong.y.zhang@intel.com>
Cc: Lv Zhiyuan <zhiyuan.lv@intel.com>
Cc: Dapeng Mi <dapeng1.mi@intel.com>
Cc: Jim Mattson <jmattson@google.com>
Acked-by: Like Xu <likexu@tencent.com>
Link: https://lore.kernel.org/r/20240307005833.827147-1-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: Explicitly disallow activatating a gfn_to_pfn_cache with INVALID_GPA
Sean Christopherson [Wed, 20 Mar 2024 00:15:42 +0000 (17:15 -0700)]
KVM: Explicitly disallow activatating a gfn_to_pfn_cache with INVALID_GPA

Explicit disallow activating a gfn_to_pfn_cache with an error gpa, i.e.
INVALID_GPA, to ensure that KVM doesn't mistake a GPA-based cache for an
HVA-based cache (KVM uses INVALID_GPA as a magic value to differentiate
between GPA-based and HVA-based caches).

WARN if KVM attempts to activate a cache with INVALID_GPA, purely so that
new caches need to at least consider what to do with a "bad" GPA, as all
existing usage of kvm_gpc_activate() guarantees gpa != INVALID_GPA.  I.e.
removing the WARN in the future is completely reasonable if doing so would
yield cleaner/better code overall.

Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
Link: https://lore.kernel.org/r/20240320001542.3203871-4-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: Check validity of offset+length of gfn_to_pfn_cache prior to activation
Sean Christopherson [Wed, 20 Mar 2024 00:15:41 +0000 (17:15 -0700)]
KVM: Check validity of offset+length of gfn_to_pfn_cache prior to activation

When activating a gfn_to_pfn_cache, verify that the offset+length is sane
and usable before marking the cache active.  Letting __kvm_gpc_refresh()
detect the problem results in a cache being marked active without setting
the GPA (or any other fields), which in turn results in KVM trying to
refresh a cache with INVALID_GPA.

Attempting to refresh a cache with INVALID_GPA isn't functionally
problematic, but it runs afoul of the sanity check that exactly one of
GPA or userspace HVA is valid, i.e. that a cache is either GPA-based or
HVA-based.

Reported-by: syzbot+106a4f72b0474e1d1b33@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/0000000000005fa5cc0613f1cebd@google.com
Fixes: 721f5b0dda78 ("KVM: pfncache: allow a cache to be activated with a fixed (userspace) HVA")
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Paul Durrant <paul@xen.org>
Reviewed-by: Paul Durrant <paul@xen.org>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Link: https://lore.kernel.org/r/20240320001542.3203871-3-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoKVM: Add helpers to consolidate gfn_to_pfn_cache's page split check
Sean Christopherson [Wed, 20 Mar 2024 00:15:40 +0000 (17:15 -0700)]
KVM: Add helpers to consolidate gfn_to_pfn_cache's page split check

Add a helper to check that the incoming length for a gfn_to_pfn_cache is
valid with respect to the cache's GPA and/or HVA.  To avoid activating a
cache with a bogus GPA, a future fix will fork the page split check in
the inner refresh path into activate() and the public rerfresh() APIs, at
which point KVM will check the length in three separate places.

Deliberately keep the "page offset" logic open coded, as the only other
path that consumes the offset, __kvm_gpc_refresh(), already needs to
differentiate between GPA-based and HVA-based caches, and it's not obvious
that using a helper is a net positive in overall code readability.

Note, for GPA-based caches, this has a subtle side effect of using the GPA
instead of the resolved HVA in the check() path, but that should be a nop
as the HVA offset is derived from the GPA, i.e. the two offsets are
identical, barring a KVM bug.

Reviewed-by: Paul Durrant <paul@xen.org>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Link: https://lore.kernel.org/r/20240320001542.3203871-2-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
14 months agoLinux 6.9-rc3
Linus Torvalds [Sun, 7 Apr 2024 20:22:46 +0000 (13:22 -0700)]
Linux 6.9-rc3

14 months agoMerge tag 'x86-urgent-2024-04-07' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 7 Apr 2024 16:33:21 +0000 (09:33 -0700)]
Merge tag 'x86-urgent-2024-04-07' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:

 - Fix MCE timer reinit locking

 - Fix/improve CoCo guest random entropy pool init

 - Fix SEV-SNP late disable bugs

 - Fix false positive objtool build warning

 - Fix header dependency bug

 - Fix resctrl CPU offlining bug

* tag 'x86-urgent-2024-04-07' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/retpoline: Add NOENDBR annotation to the SRSO dummy return thunk
  x86/mce: Make sure to grab mce_sysfs_mutex in set_bank()
  x86/CPU/AMD: Track SNP host status with cc_platform_*()
  x86/cc: Add cc_platform_set/_clear() helpers
  x86/kvm/Kconfig: Have KVM_AMD_SEV select ARCH_HAS_CC_PLATFORM
  x86/coco: Require seeding RNG with RDRAND on CoCo systems
  x86/numa/32: Include missing <asm/pgtable_areas.h>
  x86/resctrl: Fix uninitialized memory read when last CPU of domain goes offline

14 months agoMerge tag 'timers-urgent-2024-04-07' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Apr 2024 16:20:50 +0000 (09:20 -0700)]
Merge tag 'timers-urgent-2024-04-07' of git://git./linux/kernel/git/tip/tip

Pull timer fixes from Ingo Molnar:
 "Fix various timer bugs:

   - Fix a timer migration bug that may result in missed events

   - Fix timer migration group hierarchy event updates

   - Fix a PowerPC64 build warning

   - Fix a handful of DocBook annotation bugs"

* tag 'timers-urgent-2024-04-07' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  timers/migration: Return early on deactivation
  timers/migration: Fix ignored event due to missing CPU update
  vdso: Use CONFIG_PAGE_SHIFT in vdso/datapage.h
  timers: Fix text inconsistencies and spelling
  tick/sched: Fix struct tick_sched doc warnings
  tick/sched: Fix various kernel-doc warnings
  timers: Fix kernel-doc format and add Return values
  time/timekeeping: Fix kernel-doc warnings and typos
  time/timecounter: Fix inline documentation

14 months agoMerge tag 'perf-urgent-2024-04-07' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 7 Apr 2024 16:14:46 +0000 (09:14 -0700)]
Merge tag 'perf-urgent-2024-04-07' of git://git./linux/kernel/git/tip/tip

Pull x86 perf fix from Ingo Molnar:
 "Fix a combined PEBS events bug on x86 Intel CPUs"

* tag 'perf-urgent-2024-04-07' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/intel/ds: Don't clear ->pebs_data_cfg for the last PEBS event

14 months agoMerge tag 'nfsd-6.9-2' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux
Linus Torvalds [Sat, 6 Apr 2024 16:37:50 +0000 (09:37 -0700)]
Merge tag 'nfsd-6.9-2' of git://git./linux/kernel/git/cel/linux

Pull nfsd fixes from Chuck Lever:

 - Address a slow memory leak with RPC-over-TCP

 - Prevent another NFS4ERR_DELAY loop during CREATE_SESSION

* tag 'nfsd-6.9-2' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux:
  nfsd: hold a lighter-weight client reference over CB_RECALL_ANY
  SUNRPC: Fix a slow server-side memory leak with RPC-over-TCP

14 months agoMerge tag 'i2c-for-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sat, 6 Apr 2024 16:27:36 +0000 (09:27 -0700)]
Merge tag 'i2c-for-6.9-rc3' of git://git./linux/kernel/git/wsa/linux

Pull i2c fix from Wolfram Sang:
 "A host driver build fix"

* tag 'i2c-for-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: pxa: hide unused icr_bits[] variable

14 months agoMerge tag 'xfs-6.9-fixes-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Sat, 6 Apr 2024 16:14:18 +0000 (09:14 -0700)]
Merge tag 'xfs-6.9-fixes-2' of git://git./fs/xfs/xfs-linux

Pull xfs fix from Chandan Babu:

 - Allow creating new links to special files which were not associated
   with a project quota

* tag 'xfs-6.9-fixes-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: allow cross-linking special files without project quota

14 months agoMerge tag '6.9-rc2-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 6 Apr 2024 16:06:17 +0000 (09:06 -0700)]
Merge tag '6.9-rc2-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull smb client fixes from Steve French:

 - fix to retry close to avoid potential handle leaks when server
   returns EBUSY

 - DFS fixes including a fix for potential use after free

 - fscache fix

 - minor strncpy cleanup

 - reconnect race fix

 - deal with various possible UAF race conditions tearing sessions down

* tag '6.9-rc2-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  smb: client: fix potential UAF in cifs_signal_cifsd_for_reconnect()
  smb: client: fix potential UAF in smb2_is_network_name_deleted()
  smb: client: fix potential UAF in is_valid_oplock_break()
  smb: client: fix potential UAF in smb2_is_valid_oplock_break()
  smb: client: fix potential UAF in smb2_is_valid_lease_break()
  smb: client: fix potential UAF in cifs_stats_proc_show()
  smb: client: fix potential UAF in cifs_stats_proc_write()
  smb: client: fix potential UAF in cifs_dump_full_key()
  smb: client: fix potential UAF in cifs_debug_files_proc_show()
  smb3: retrying on failed server close
  smb: client: serialise cifs_construct_tcon() with cifs_mount_mutex
  smb: client: handle DFS tcons in cifs_construct_tcon()
  smb: client: refresh referral without acquiring refpath_lock
  smb: client: guarantee refcounted children from parent session
  cifs: Fix caching to try to do open O_WRONLY as rdwr on server
  smb: client: fix UAF in smb2_reconnect_server()
  smb: client: replace deprecated strncpy with strscpy

14 months agox86/retpoline: Add NOENDBR annotation to the SRSO dummy return thunk
Borislav Petkov (AMD) [Fri, 5 Apr 2024 14:46:37 +0000 (16:46 +0200)]
x86/retpoline: Add NOENDBR annotation to the SRSO dummy return thunk

srso_alias_untrain_ret() is special code, even if it is a dummy
which is called in the !SRSO case, so annotate it like its real
counterpart, to address the following objtool splat:

  vmlinux.o: warning: objtool: .export_symbol+0x2b290: data relocation to !ENDBR: srso_alias_untrain_ret+0x0

Fixes: 4535e1a4174c ("x86/bugs: Fix the SRSO mitigation on Zen3/4")
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: https://lore.kernel.org/r/20240405144637.17908-1-bp@kernel.org
14 months agoMerge branch 'linus' into x86/urgent, to pick up dependent commit
Ingo Molnar [Sat, 6 Apr 2024 11:00:32 +0000 (13:00 +0200)]
Merge branch 'linus' into x86/urgent, to pick up dependent commit

We want to fix:

  0e110732473e ("x86/retpoline: Do the necessary fixup to the Zen3/4 srso return thunk for !SRSO")

So merge in Linus's latest into x86/urgent to have it available.

Signed-off-by: Ingo Molnar <mingo@kernel.org>
14 months agoMerge tag 'i2c-host-fixes-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Wolfram Sang [Sat, 6 Apr 2024 09:29:15 +0000 (11:29 +0200)]
Merge tag 'i2c-host-fixes-6.9-rc3' of git://git./linux/kernel/git/andi.shyti/linux into i2c/for-current

An unused const variable kind of error has been fixed by placing
the definition of icr_bits[] inside the ifdef block where it is
used.

14 months agoMerge tag 'firewire-fixes-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 6 Apr 2024 04:25:31 +0000 (21:25 -0700)]
Merge tag 'firewire-fixes-6.9-rc2' of git://git./linux/kernel/git/ieee1394/linux1394

Pull firewire fixes from Takashi Sakamoto:
 "The firewire-ohci kernel module has a parameter for verbose kernel
  logging. It is well-known that it logs the spurious IRQ for bus-reset
  event due to the unmasked register for IRQ event. This update fixes
  the issue"

* tag 'firewire-fixes-6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394:
  firewire: ohci: mask bus reset interrupts between ISR and bottom half

14 months agofirewire: ohci: mask bus reset interrupts between ISR and bottom half
Adam Goldman [Sun, 24 Mar 2024 22:38:41 +0000 (07:38 +0900)]
firewire: ohci: mask bus reset interrupts between ISR and bottom half

In the FireWire OHCI interrupt handler, if a bus reset interrupt has
occurred, mask bus reset interrupts until bus_reset_work has serviced and
cleared the interrupt.

Normally, we always leave bus reset interrupts masked. We infer the bus
reset from the self-ID interrupt that happens shortly thereafter. A
scenario where we unmask bus reset interrupts was introduced in 2008 in
a007bb857e0b26f5d8b73c2ff90782d9c0972620: If
OHCI_PARAM_DEBUG_BUSRESETS (8) is set in the debug parameter bitmask, we
will unmask bus reset interrupts so we can log them.

irq_handler logs the bus reset interrupt. However, we can't clear the bus
reset event flag in irq_handler, because we won't service the event until
later. irq_handler exits with the event flag still set. If the
corresponding interrupt is still unmasked, the first bus reset will
usually freeze the system due to irq_handler being called again each
time it exits. This freeze can be reproduced by loading firewire_ohci
with "modprobe firewire_ohci debug=-1" (to enable all debugging output).
Apparently there are also some cases where bus_reset_work will get called
soon enough to clear the event, and operation will continue normally.

This freeze was first reported a few months after a007bb85 was committed,
but until now it was never fixed. The debug level could safely be set
to -1 through sysfs after the module was loaded, but this would be
ineffectual in logging bus reset interrupts since they were only
unmasked during initialization.

irq_handler will now leave the event flag set but mask bus reset
interrupts, so irq_handler won't be called again and there will be no
freeze. If OHCI_PARAM_DEBUG_BUSRESETS is enabled, bus_reset_work will
unmask the interrupt after servicing the event, so future interrupts
will be caught as desired.

As a side effect to this change, OHCI_PARAM_DEBUG_BUSRESETS can now be
enabled through sysfs in addition to during initial module loading.
However, when enabled through sysfs, logging of bus reset interrupts will
be effective only starting with the second bus reset, after
bus_reset_work has executed.

Signed-off-by: Adam Goldman <adamg@pobox.com>
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
14 months agoMerge tag 'spi-fix-v6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi
Linus Torvalds [Sat, 6 Apr 2024 00:26:43 +0000 (17:26 -0700)]
Merge tag 'spi-fix-v6.9-rc2' of git://git./linux/kernel/git/broonie/spi

Pull spi fixes from Mark Brown:
 "A few small driver specific fixes, the most important being the
  s3c64xx change which is likely to be hit during normal operation"

* tag 'spi-fix-v6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
  spi: mchp-pci1xxx: Fix a possible null pointer dereference in pci1xxx_spi_probe
  spi: spi-fsl-lpspi: remove redundant spi_controller_put call
  spi: s3c64xx: Use DMA mode from fifo size

14 months agoMerge tag 'regulator-fix-v6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 6 Apr 2024 00:24:04 +0000 (17:24 -0700)]
Merge tag 'regulator-fix-v6.9-rc2' of git://git./linux/kernel/git/broonie/regulator

Pull regulator fix from Mark Brown:
 "One simple regualtor fix, fixing module autoloading on tps65132"

* tag 'regulator-fix-v6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: tps65132: Add of_match table

14 months agoMerge tag 'regmap-fix-v6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 6 Apr 2024 00:21:16 +0000 (17:21 -0700)]
Merge tag 'regmap-fix-v6.9-rc2' of git://git./linux/kernel/git/broonie/regmap

Pull regmap fixes from Mark Brown:
 "Richard found a nasty corner case in the maple tree code which he
  fixed, and also fixed a compiler warning which was showing up with the
  toolchain he uses and helpfully identified a possible incorrect error
  code which could have runtime impacts"

* tag 'regmap-fix-v6.9-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
  regmap: maple: Fix uninitialized symbol 'ret' warnings
  regmap: maple: Fix cache corruption in regcache_maple_drop()

14 months agoMerge tag 'block-6.9-20240405' of git://git.kernel.dk/linux
Linus Torvalds [Sat, 6 Apr 2024 00:04:11 +0000 (17:04 -0700)]
Merge tag 'block-6.9-20240405' of git://git.kernel.dk/linux

Pull block fixes from Jens Axboe:

 - NVMe pull request via Keith:
      - Atomic queue limits fixes (Christoph)
      - Fabrics fixes (Hannes, Daniel)

 - Discard overflow fix (Li)

 - Cleanup fix for null_blk (Damien)

* tag 'block-6.9-20240405' of git://git.kernel.dk/linux:
  nvme-fc: rename free_ctrl callback to match name pattern
  nvmet-fc: move RCU read lock to nvmet_fc_assoc_exists
  nvmet: implement unique discovery NQN
  nvme: don't create a multipath node for zero capacity devices
  nvme: split nvme_update_zone_info
  nvme-multipath: don't inherit LBA-related fields for the multipath node
  block: fix overflow in blk_ioctl_discard()
  nullblk: Fix cleanup order in null_add_dev() error path

14 months agoMerge tag 'io_uring-6.9-20240405' of git://git.kernel.dk/linux
Linus Torvalds [Fri, 5 Apr 2024 23:58:52 +0000 (16:58 -0700)]
Merge tag 'io_uring-6.9-20240405' of git://git.kernel.dk/linux

Pull io_uring fixes from Jens Axboe:

 - Backport of some fixes that came up during development of the 6.10
   io_uring patches. This includes some kbuf cleanups and reference
   fixes.

 - Disable multishot read if we don't have NOWAIT support on the target

 - Fix for a dependency issue with workqueue flushing

* tag 'io_uring-6.9-20240405' of git://git.kernel.dk/linux:
  io_uring/kbuf: hold io_buffer_list reference over mmap
  io_uring/kbuf: protect io_buffer_list teardown with a reference
  io_uring/kbuf: get rid of bl->is_ready
  io_uring/kbuf: get rid of lower BGID lists
  io_uring: use private workqueue for exit work
  io_uring: disable io-wq execution of multishot NOWAIT requests
  io_uring/rw: don't allow multishot reads without NOWAIT support

14 months agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Fri, 5 Apr 2024 23:54:54 +0000 (16:54 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "The most important is the libsas fix, which is a problem for DMA to a
  kmalloc'd structure too small causing cache line interference. The
  other fixes (all in drivers) are mostly for allocation length fixes,
  error leg unwinding, suspend races and a missing retry"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: ufs: core: Fix MCQ mode dev command timeout
  scsi: libsas: Align SMP request allocation to ARCH_DMA_MINALIGN
  scsi: sd: Unregister device if device_add_disk() failed in sd_probe()
  scsi: ufs: core: WLUN suspend dev/link state error recovery
  scsi: mylex: Fix sysfs buffer lengths

14 months agoMerge tag 'devicetree-fixes-for-6.9-1' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 5 Apr 2024 21:07:22 +0000 (14:07 -0700)]
Merge tag 'devicetree-fixes-for-6.9-1' of git://git./linux/kernel/git/robh/linux

Pull devicetree fixes from Rob Herring:

 - Fix NIOS2 boot with external DTB

 - Add missing synchronization needed between fw_devlink and DT overlay
   removals

 - Fix some unit-address regex's to be hex only

 - Drop some 10+ year old "unstable binding" statements

 - Add new SoCs to QCom UFS binding

 - Add TPM bindings to TPM maintainers

* tag 'devicetree-fixes-for-6.9-1' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  nios2: Only use built-in devicetree blob if configured to do so
  dt-bindings: timer: narrow regex for unit address to hex numbers
  dt-bindings: soc: fsl: narrow regex for unit address to hex numbers
  dt-bindings: remoteproc: ti,davinci: remove unstable remark
  dt-bindings: clock: ti: remove unstable remark
  dt-bindings: clock: keystone: remove unstable remark
  of: module: prevent NULL pointer dereference in vsnprintf()
  dt-bindings: ufs: qcom: document SM6125 UFS
  dt-bindings: ufs: qcom: document SC7180 UFS
  dt-bindings: ufs: qcom: document SC8180X UFS
  of: dynamic: Synchronize of_changeset_destroy() with the devlink removals
  driver core: Introduce device_link_wait_removal()
  docs: dt-bindings: add missing address/size-cells to example
  MAINTAINERS: Add TPM DT bindings to TPM maintainers

14 months agoMerge tag 'mm-hotfixes-stable-2024-04-05-11-30' of git://git.kernel.org/pub/scm/linux...
Linus Torvalds [Fri, 5 Apr 2024 20:30:01 +0000 (13:30 -0700)]
Merge tag 'mm-hotfixes-stable-2024-04-05-11-30' of git://git./linux/kernel/git/akpm/mm

Pull misc fixes from Andrew Morton:
 "8 hotfixes, 3 are cc:stable

  There are a couple of fixups for this cycle's vmalloc changes and one
  for the stackdepot changes. And a fix for a very old x86 PAT issue
  which can cause a warning splat"

* tag 'mm-hotfixes-stable-2024-04-05-11-30' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
  stackdepot: rename pool_index to pool_index_plus_1
  x86/mm/pat: fix VM_PAT handling in COW mappings
  MAINTAINERS: change vmware.com addresses to broadcom.com
  selftests/mm: include strings.h for ffsl
  mm: vmalloc: fix lockdep warning
  mm: vmalloc: bail out early in find_vmap_area() if vmap is not init
  init: open output files from cpio unpacking with O_LARGEFILE
  mm/secretmem: fix GUP-fast succeeding on secretmem folios

14 months agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 5 Apr 2024 20:12:35 +0000 (13:12 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fix from Catalin Marinas:
 "arm64/ptrace fix to use the correct SVE layout based on the saved
  floating point state rather than the TIF_SVE flag. The latter may be
  left on during syscalls even if the SVE state is discarded"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64/ptrace: Use saved floating point state type to determine SVE layout

14 months agoMerge tag 'riscv-for-linus-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 5 Apr 2024 20:09:48 +0000 (13:09 -0700)]
Merge tag 'riscv-for-linus-6.9-rc3' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Palmer Dabbelt:

 - A fix for an __{get,put}_kernel_nofault to avoid an uninitialized
   value causing spurious failures

 - compat_vdso.so.dbg is now installed to the standard install location

 - A fix to avoid initializing PERF_SAMPLE_BRANCH_*-related events, as
   they aren't supported and will just later fail

 - A fix to make AT_VECTOR_SIZE_ARCH correct now that we're providing
   AT_MINSIGSTKSZ

 - pgprot_nx() is now implemented, which fixes vmap W^X protection

 - A fix for the vector save/restore code, which at least manifests as
   corrupted vector state when a signal is taken

 - A fix for a race condition in instruction patching

 - A fix to avoid leaking the kernel-mode GP to userspace, which is a
   kernel pointer leak that can be used to defeat KASLR in various ways

 - A handful of smaller fixes to build warnings, an overzealous printk,
   and some missing tracing annotations

* tag 'riscv-for-linus-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: process: Fix kernel gp leakage
  riscv: Disable preemption when using patch_map()
  riscv: Fix warning by declaring arch_cpu_idle() as noinstr
  riscv: use KERN_INFO in do_trap
  riscv: Fix vector state restore in rt_sigreturn()
  riscv: mm: implement pgprot_nx
  riscv: compat_vdso: align VDSOAS build log
  RISC-V: Update AT_VECTOR_SIZE_ARCH for new AT_MINSIGSTKSZ
  riscv: Mark __se_sys_* functions __used
  drivers/perf: riscv: Disable PERF_SAMPLE_BRANCH_* while not supported
  riscv: compat_vdso: install compat_vdso.so.dbg to /lib/modules/*/vdso/
  riscv: hwprobe: do not produce frtace relocation
  riscv: Fix spurious errors from __get/put_kernel_nofault
  riscv: mm: Fix prototype to avoid discarding const

14 months agoMerge tag 's390-6.9-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Fri, 5 Apr 2024 20:07:25 +0000 (13:07 -0700)]
Merge tag 's390-6.9-3' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Alexander Gordeev:

 - Fix missing NULL pointer check when determining guest/host fault

 - Mark all functions in asm/atomic_ops.h, asm/atomic.h and
   asm/preempt.h as __always_inline to avoid unwanted instrumentation

 - Fix removal of a Processor Activity Instrumentation (PAI) sampling
   event in PMU device driver

 - Align system call table on 8 bytes

* tag 's390-6.9-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/entry: align system call table on 8 bytes
  s390/pai: fix sampling event removal for PMU device driver
  s390/preempt: mark all functions __always_inline
  s390/atomic: mark all functions __always_inline
  s390/mm: fix NULL pointer dereference

14 months agoMerge tag 'pm-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Fri, 5 Apr 2024 19:55:40 +0000 (12:55 -0700)]
Merge tag 'pm-6.9-rc3' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fix from Rafael Wysocki:
 "Fix a recent Energy Model change that went against a recent scheduler
  change made independently (Vincent Guittot)"

* tag 'pm-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  PM: EM: fix wrong utilization estimation in em_cpu_energy()

14 months agoMerge tag 'thermal-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 5 Apr 2024 19:51:32 +0000 (12:51 -0700)]
Merge tag 'thermal-6.9-rc3' of git://git./linux/kernel/git/rafael/linux-pm

Pull thermal control fixes from Rafael Wysocki:
 "These fix two power allocator thermal governor issues and an ACPI
  thermal driver regression that all were introduced during the 6.8
  development cycle.

  Specifics:

   - Allow the power allocator thermal governor to bind to a thermal
     zone without cooling devices and/or without trip points (Nikita
     Travkin)

   - Make the ACPI thermal driver register a tripless thermal zone when
     it cannot find any usable trip points instead of returning an error
     from acpi_thermal_add() (Stephen Horvath)"

* tag 'thermal-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  thermal: gov_power_allocator: Allow binding without trip points
  thermal: gov_power_allocator: Allow binding without cooling devices
  ACPI: thermal: Register thermal zones without valid trip points

14 months agoMerge tag 'gpio-fixes-for-v6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 5 Apr 2024 19:12:19 +0000 (12:12 -0700)]
Merge tag 'gpio-fixes-for-v6.9-rc3' of git://git./linux/kernel/git/brgl/linux

Pull gpio fixes from Bartosz Golaszewski:

 - make sure GPIO devices are registered with the subsystem before
   trying to return them to a caller of gpio_device_find()

 - fix two issues with incorrect sanitization of the interrupt labels

* tag 'gpio-fixes-for-v6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux:
  gpio: cdev: fix missed label sanitizing in debounce_setup()
  gpio: cdev: check for NULL labels when sanitizing them for irqs
  gpiolib: Fix triggering "kobject: 'gpiochipX' is not initialized, yet" kobject_get() errors

14 months agoMerge tag 'ata-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/libata/linux
Linus Torvalds [Fri, 5 Apr 2024 19:09:16 +0000 (12:09 -0700)]
Merge tag 'ata-6.9-rc3' of git://git./linux/kernel/git/libata/linux

Pull ata fixes from Damien Le Moal:

 - Compilation warning fixes from Arnd: one in the sata_sx4 driver due
   to an incorrect calculation of the parameters passed to memcpy() and
   another one in the sata_mv driver when CONFIG_PCI is not set

 - Drop the owner driver field assignment in the pata_macio driver. That
   is not needed as the PCI core code does that already (Krzysztof)

 - Remove an unusued field in struct st_ahci_drv_data of the ahci_st
   driver (Christophe)

 - Add a missing clock probe error check in the sata_gemini driver
   (Chen)

* tag 'ata-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/libata/linux:
  ata: sata_gemini: Check clk_enable() result
  ata: sata_mv: Fix PCI device ID table declaration compilation warning
  ata: ahci_st: Remove an unused field in struct st_ahci_drv_data
  ata: pata_macio: drop driver owner assignment
  ata: sata_sx4: fix pdc20621_get_from_dimm() on 64-bit

14 months agoMerge tag 'sound-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 5 Apr 2024 18:58:55 +0000 (11:58 -0700)]
Merge tag 'sound-6.9-rc3' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "This became a bit bigger collection of patches, but almost all are
  about device-specific fixes, and should be safe for 6.9:

   - Lots of ASoC Intel SOF-related fixes/updates

   - Locking fixes in SoundWire drivers

   - ASoC AMD ACP/SOF updates

   - ASoC ES8326 codec fixes

   - HD-audio codec fixes and quirks

   - A regression fix in emu10k1 synth code"

* tag 'sound-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound: (49 commits)
  ASoC: SOF: Core: Add remove_late() to sof_init_environment failure path
  ASoC: SOF: amd: fix for false dsp interrupts
  ASoC: SOF: Intel: lnl: Disable DMIC/SSP offload on remove
  ASoC: Intel: avs: boards: Add modules description
  ASoC: codecs: ES8326: Removing the control of ADC_SCALE
  ASoC: codecs: ES8326: Solve a headphone detection issue after suspend and resume
  ASoC: codecs: ES8326: modify clock table
  ASoC: codecs: ES8326: Solve error interruption issue
  ALSA: line6: Zero-initialize message buffers
  ALSA: hda/realtek: cs35l41: Support ASUS ROG G634JYR
  ALSA: hda/realtek: Update Panasonic CF-SZ6 quirk to support headset with microphone
  ALSA: hda/realtek: Add sound quirks for Lenovo Legion slim 7 16ARHA7 models
  Revert "ALSA: emu10k1: fix synthesizer sample playback position and caching"
  OSS: dmasound/paula: Mark driver struct with __refdata to prevent section mismatch
  ALSA: hda/realtek: Add quirks for ASUS Laptops using CS35L56
  ASoC: amd: acp: fix for acp_init function error handling
  ASoC: tas2781: mark dvc_tlv with __maybe_unused
  ASoC: ops: Fix wraparound for mask in snd_soc_get_volsw
  ASoC: rt-sdw*: add __func__ to all error logs
  ASoC: rt722-sdca-sdw: fix locking sequence
  ...

14 months agoMerge tag 'drm-fixes-2024-04-05' of https://gitlab.freedesktop.org/drm/kernel
Linus Torvalds [Fri, 5 Apr 2024 18:53:46 +0000 (11:53 -0700)]
Merge tag 'drm-fixes-2024-04-05' of https://gitlab.freedesktop.org/drm/kernel

Pull drm fixes from Dave Airlie:
 "Weekly fixes, mostly xe and i915, amdgpu on a week off, otherwise a
  nouveau fix for a crash with new vulkan cts tests, and a couple of
  cleanups and misc fixes.

  display:
   - fix typos in kerneldoc

  prime:
   - unbreak dma-buf export for virt-gpu

  nouveau:
   - uvmm: fix remap address calculation
   - minor cleanups

  panfrost:
   - fix power-transition timeouts

  xe:
   - Stop using system_unbound_wq for preempt fences
   - Fix saving unordered rebinding fences by attaching them as kernel
     feces to the vm's resv
   - Fix TLB invalidation fences completing out of order
   - Move rebind TLB invalidation to the ring ops to reduce the latency

  i915:
   - A few DisplayPort related fixes
   - eDP PSR fixes
   - Remove some VM space restrictions on older platforms
   - Disable automatic load CCS load balancing"

* tag 'drm-fixes-2024-04-05' of https://gitlab.freedesktop.org/drm/kernel: (22 commits)
  drm/xe: Use ordered wq for preempt fence waiting
  drm/xe: Move vma rebinding to the drm_exec locking loop
  drm/xe: Make TLB invalidation fences unordered
  drm/xe: Rework rebinding
  drm/xe: Use ring ops TLB invalidation for rebinds
  drm/i915/mst: Reject FEC+MST on ICL
  drm/i915/mst: Limit MST+DSC to TGL+
  drm/i915/dp: Fix the computation for compressed_bpp for DISPLAY < 13
  drm/i915/gt: Enable only one CCS for compute workload
  drm/i915/gt: Do not generate the command streamer for all the CCS
  drm/i915/gt: Disable HW load balancing for CCS
  drm/i915/gt: Limit the reserved VM space to only the platforms that need it
  drm/i915/psr: Fix intel_psr2_sel_fetch_et_alignment usage
  drm/i915/psr: Move writing early transport pipe src
  drm/i915/psr: Calculate PIPE_SRCSZ_ERLY_TPT value
  drm/i915/dp: Remove support for UHBR13.5
  drm/i915/dp: Fix DSC state HW readout for SST connectors
  drm/display: fix typo
  drm/prime: Unbreak virtgpu dma-buf export
  nouveau/uvmm: fix addr/range calcs for remap operations
  ...

14 months agostackdepot: rename pool_index to pool_index_plus_1
Peter Collingbourne [Tue, 2 Apr 2024 00:14:58 +0000 (17:14 -0700)]
stackdepot: rename pool_index to pool_index_plus_1

Commit 3ee34eabac2a ("lib/stackdepot: fix first entry having a 0-handle")
changed the meaning of the pool_index field to mean "the pool index plus
1".  This made the code accessing this field less self-documenting, as
well as causing debuggers such as drgn to not be able to easily remain
compatible with both old and new kernels, because they typically do that
by testing for presence of the new field.  Because stackdepot is a
debugging tool, we should make sure that it is debugger friendly.
Therefore, give the field a different name to improve readability as well
as enabling debugger backwards compatibility.

This is needed in 6.9, which would otherwise become an odd release with
the new semantics and old name so debuggers wouldn't recognize the new
semantics there.

Fixes: 3ee34eabac2a ("lib/stackdepot: fix first entry having a 0-handle")
Link: https://lkml.kernel.org/r/20240402001500.53533-1-pcc@google.com
Link: https://linux-review.googlesource.com/id/Ib3e70c36c1d230dd0a118dc22649b33e768b9f88
Signed-off-by: Peter Collingbourne <pcc@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Alexander Potapenko <glider@google.com>
Acked-by: Marco Elver <elver@google.com>
Acked-by: Oscar Salvador <osalvador@suse.de>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Omar Sandoval <osandov@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
14 months agox86/mm/pat: fix VM_PAT handling in COW mappings
David Hildenbrand [Wed, 3 Apr 2024 21:21:30 +0000 (23:21 +0200)]
x86/mm/pat: fix VM_PAT handling in COW mappings

PAT handling won't do the right thing in COW mappings: the first PTE (or,
in fact, all PTEs) can be replaced during write faults to point at anon
folios.  Reliably recovering the correct PFN and cachemode using
follow_phys() from PTEs will not work in COW mappings.

Using follow_phys(), we might just get the address+protection of the anon
folio (which is very wrong), or fail on swap/nonswap entries, failing
follow_phys() and triggering a WARN_ON_ONCE() in untrack_pfn() and
track_pfn_copy(), not properly calling free_pfn_range().

In free_pfn_range(), we either wouldn't call memtype_free() or would call
it with the wrong range, possibly leaking memory.

To fix that, let's update follow_phys() to refuse returning anon folios,
and fallback to using the stored PFN inside vma->vm_pgoff for COW mappings
if we run into that.

We will now properly handle untrack_pfn() with COW mappings, where we
don't need the cachemode.  We'll have to fail fork()->track_pfn_copy() if
the first page was replaced by an anon folio, though: we'd have to store
the cachemode in the VMA to make this work, likely growing the VMA size.

For now, lets keep it simple and let track_pfn_copy() just fail in that
case: it would have failed in the past with swap/nonswap entries already,
and it would have done the wrong thing with anon folios.

Simple reproducer to trigger the WARN_ON_ONCE() in untrack_pfn():

<--- C reproducer --->
 #include <stdio.h>
 #include <sys/mman.h>
 #include <unistd.h>
 #include <liburing.h>

 int main(void)
 {
         struct io_uring_params p = {};
         int ring_fd;
         size_t size;
         char *map;

         ring_fd = io_uring_setup(1, &p);
         if (ring_fd < 0) {
                 perror("io_uring_setup");
                 return 1;
         }
         size = p.sq_off.array + p.sq_entries * sizeof(unsigned);

         /* Map the submission queue ring MAP_PRIVATE */
         map = mmap(0, size, PROT_READ | PROT_WRITE, MAP_PRIVATE,
                    ring_fd, IORING_OFF_SQ_RING);
         if (map == MAP_FAILED) {
                 perror("mmap");
                 return 1;
         }

         /* We have at least one page. Let's COW it. */
         *map = 0;
         pause();
         return 0;
 }
<--- C reproducer --->

On a system with 16 GiB RAM and swap configured:
 # ./iouring &
 # memhog 16G
 # killall iouring
[  301.552930] ------------[ cut here ]------------
[  301.553285] WARNING: CPU: 7 PID: 1402 at arch/x86/mm/pat/memtype.c:1060 untrack_pfn+0xf4/0x100
[  301.553989] Modules linked in: binfmt_misc nft_fib_inet nft_fib_ipv4 nft_fib_ipv6 nft_fib nft_reject_g
[  301.558232] CPU: 7 PID: 1402 Comm: iouring Not tainted 6.7.5-100.fc38.x86_64 #1
[  301.558772] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.3-0-ga6ed6b701f0a-prebu4
[  301.559569] RIP: 0010:untrack_pfn+0xf4/0x100
[  301.559893] Code: 75 c4 eb cf 48 8b 43 10 8b a8 e8 00 00 00 3b 6b 28 74 b8 48 8b 7b 30 e8 ea 1a f7 000
[  301.561189] RSP: 0018:ffffba2c0377fab8 EFLAGS: 00010282
[  301.561590] RAX: 00000000ffffffea RBX: ffff9208c8ce9cc0 RCX: 000000010455e047
[  301.562105] RDX: 07fffffff0eb1e0a RSI: 0000000000000000 RDI: ffff9208c391d200
[  301.562628] RBP: 0000000000000000 R08: ffffba2c0377fab8 R09: 0000000000000000
[  301.563145] R10: ffff9208d2292d50 R11: 0000000000000002 R12: 00007fea890e0000
[  301.563669] R13: 0000000000000000 R14: ffffba2c0377fc08 R15: 0000000000000000
[  301.564186] FS:  0000000000000000(0000) GS:ffff920c2fbc0000(0000) knlGS:0000000000000000
[  301.564773] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  301.565197] CR2: 00007fea88ee8a20 CR3: 00000001033a8000 CR4: 0000000000750ef0
[  301.565725] PKRU: 55555554
[  301.565944] Call Trace:
[  301.566148]  <TASK>
[  301.566325]  ? untrack_pfn+0xf4/0x100
[  301.566618]  ? __warn+0x81/0x130
[  301.566876]  ? untrack_pfn+0xf4/0x100
[  301.567163]  ? report_bug+0x171/0x1a0
[  301.567466]  ? handle_bug+0x3c/0x80
[  301.567743]  ? exc_invalid_op+0x17/0x70
[  301.568038]  ? asm_exc_invalid_op+0x1a/0x20
[  301.568363]  ? untrack_pfn+0xf4/0x100
[  301.568660]  ? untrack_pfn+0x65/0x100
[  301.568947]  unmap_single_vma+0xa6/0xe0
[  301.569247]  unmap_vmas+0xb5/0x190
[  301.569532]  exit_mmap+0xec/0x340
[  301.569801]  __mmput+0x3e/0x130
[  301.570051]  do_exit+0x305/0xaf0
...

Link: https://lkml.kernel.org/r/20240403212131.929421-3-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reported-by: Wupeng Ma <mawupeng1@huawei.com>
Closes: https://lkml.kernel.org/r/20240227122814.3781907-1-mawupeng1@huawei.com
Fixes: b1a86e15dc03 ("x86, pat: remove the dependency on 'vm_pgoff' in track/untrack pfn vma routines")
Fixes: 5899329b1910 ("x86: PAT: implement track/untrack of pfnmap regions for x86 - v3")
Acked-by: Ingo Molnar <mingo@kernel.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Borislav Petkov <bp@alien8.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
14 months agoMAINTAINERS: change vmware.com addresses to broadcom.com
Alexey Makhalov [Tue, 2 Apr 2024 23:23:34 +0000 (16:23 -0700)]
MAINTAINERS: change vmware.com addresses to broadcom.com

Update all remaining vmware.com email addresses to actual broadcom.com.

Add corresponding .mailmap entries for maintainers who contributed in the
past as the vmware.com address will start bouncing soon.

Maintainership update. Jeff Sipek has left VMware, Nick Shi will be
maintaining VMware PTP.

Link: https://lkml.kernel.org/r/20240402232334.33167-1-alexey.makhalov@broadcom.com
Signed-off-by: Alexey Makhalov <alexey.makhalov@broadcom.com>
Acked-by: Florian Fainelli <florian.fainelli@broadcom.com>
Acked-by: Ajay Kaher <ajay.kaher@broadcom.com>
Acked-by: Ronak Doshi <ronak.doshi@broadcom.com>
Acked-by: Nick Shi <nick.shi@broadcom.com>
Acked-by: Bryan Tan <bryan-bt.tan@broadcom.com>
Acked-by: Vishnu Dasa <vishnu.dasa@broadcom.com>
Acked-by: Vishal Bhakta <vishal.bhakta@broadcom.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
14 months agoselftests/mm: include strings.h for ffsl
Edward Liaw [Fri, 29 Mar 2024 18:58:10 +0000 (18:58 +0000)]
selftests/mm: include strings.h for ffsl

Got a compilation error on Android for ffsl after 91b80cc5b39f
("selftests: mm: fix map_hugetlb failure on 64K page size systems")
included vm_util.h.

Link: https://lkml.kernel.org/r/20240329185814.16304-1-edliaw@google.com
Fixes: af605d26a8f2 ("selftests/mm: merge util.h into vm_util.h")
Signed-off-by: Edward Liaw <edliaw@google.com>
Reviewed-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: "Mike Rapoport (IBM)" <rppt@kernel.org>
Cc: Peter Xu <peterx@redhat.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
14 months agomm: vmalloc: fix lockdep warning
Uladzislau Rezki (Sony) [Thu, 28 Mar 2024 14:03:30 +0000 (15:03 +0100)]
mm: vmalloc: fix lockdep warning

A lockdep reports a possible deadlock in the find_vmap_area_exceed_addr_lock()
function:

============================================
WARNING: possible recursive locking detected
6.9.0-rc1-00060-ged3ccc57b108-dirty #6140 Not tainted
--------------------------------------------
drgn/455 is trying to acquire lock:
ffff0000c00131d0 (&vn->busy.lock/1){+.+.}-{2:2}, at: find_vmap_area_exceed_addr_lock+0x64/0x124

but task is already holding lock:
ffff0000c0011878 (&vn->busy.lock/1){+.+.}-{2:2}, at: find_vmap_area_exceed_addr_lock+0x64/0x124

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(&vn->busy.lock/1);
  lock(&vn->busy.lock/1);

 *** DEADLOCK ***

indeed it can happen if the find_vmap_area_exceed_addr_lock() gets called
concurrently because it tries to acquire two nodes locks.  It was done to
prevent removing a lowest VA found on a previous step.

To address this a lowest VA is found first without holding a node lock
where it resides.  As a last step we check if a VA still there because it
can go away, if removed, proceed with next lowest.

[akpm@linux-foundation.org: fix comment typos, per Baoquan]
Link: https://lkml.kernel.org/r/20240328140330.4747-1-urezki@gmail.com
Fixes: 53becf32aec1 ("mm: vmalloc: support multiple nodes in vread_iter")
Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Tested-by: Jens Axboe <axboe@kernel.dk>
Tested-by: Omar Sandoval <osandov@fb.com>
Reported-by: Jens Axboe <axboe@kernel.dk>
Cc: Baoquan He <bhe@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Lorenzo Stoakes <lstoakes@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sony.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
14 months agomm: vmalloc: bail out early in find_vmap_area() if vmap is not init
Uladzislau Rezki (Sony) [Sat, 23 Mar 2024 14:15:44 +0000 (15:15 +0100)]
mm: vmalloc: bail out early in find_vmap_area() if vmap is not init

During the boot the s390 system triggers "spinlock bad magic" messages
if the spinlock debugging is enabled:

[    0.465445] BUG: spinlock bad magic on CPU#0, swapper/0
[    0.465490]  lock: single+0x1860/0x1958, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
[    0.466067] CPU: 0 PID: 0 Comm: swapper Not tainted 6.8.0-12955-g8e938e398669 #1
[    0.466188] Hardware name: QEMU 8561 QEMU (KVM/Linux)
[    0.466270] Call Trace:
[    0.466470]  [<00000000011f26c8>] dump_stack_lvl+0x98/0xd8
[    0.466516]  [<00000000001dcc6a>] do_raw_spin_lock+0x8a/0x108
[    0.466545]  [<000000000042146c>] find_vmap_area+0x6c/0x108
[    0.466572]  [<000000000042175a>] find_vm_area+0x22/0x40
[    0.466597]  [<000000000012f152>] __set_memory+0x132/0x150
[    0.466624]  [<0000000001cc0398>] vmem_map_init+0x40/0x118
[    0.466651]  [<0000000001cc0092>] paging_init+0x22/0x68
[    0.466677]  [<0000000001cbbed2>] setup_arch+0x52a/0x708
[    0.466702]  [<0000000001cb6140>] start_kernel+0x80/0x5c8
[    0.466727]  [<0000000000100036>] startup_continue+0x36/0x40

it happens because such system tries to access some vmap areas
whereas the vmalloc initialization is not even yet done:

[    0.465490] lock: single+0x1860/0x1958, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
[    0.466067] CPU: 0 PID: 0 Comm: swapper Not tainted 6.8.0-12955-g8e938e398669 #1
[    0.466188] Hardware name: QEMU 8561 QEMU (KVM/Linux)
[    0.466270] Call Trace:
[    0.466470] dump_stack_lvl (lib/dump_stack.c:117)
[    0.466516] do_raw_spin_lock (kernel/locking/spinlock_debug.c:87 kernel/locking/spinlock_debug.c:115)
[    0.466545] find_vmap_area (mm/vmalloc.c:1059 mm/vmalloc.c:2364)
[    0.466572] find_vm_area (mm/vmalloc.c:3150)
[    0.466597] __set_memory (arch/s390/mm/pageattr.c:360 arch/s390/mm/pageattr.c:393)
[    0.466624] vmem_map_init (./arch/s390/include/asm/set_memory.h:55 arch/s390/mm/vmem.c:660)
[    0.466651] paging_init (arch/s390/mm/init.c:97)
[    0.466677] setup_arch (arch/s390/kernel/setup.c:972)
[    0.466702] start_kernel (init/main.c:899)
[    0.466727] startup_continue (arch/s390/kernel/head64.S:35)
[    0.466811] INFO: lockdep is turned off.
...
[    0.718250] vmalloc init - busy lock init 0000000002871860
[    0.718328] vmalloc init - busy lock init 00000000028731b8

Some background. It worked before because the lock that is in question
was statically defined and initialized. As of now, the locks and data
structures are initialized in the vmalloc_init() function.

To address that issue add the check whether the "vmap_initialized"
variable is set, if not find_vmap_area() bails out on entry returning NULL.

Link: https://lkml.kernel.org/r/20240323141544.4150-1-urezki@gmail.com
Fixes: 72210662c5a2 ("mm: vmalloc: offload free_vmap_area_lock lock")
Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Baoquan He <bhe@redhat.com>
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Lorenzo Stoakes <lstoakes@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sony.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
14 months agoinit: open output files from cpio unpacking with O_LARGEFILE
John Sperbeck [Sat, 23 Mar 2024 15:29:34 +0000 (08:29 -0700)]
init: open output files from cpio unpacking with O_LARGEFILE

If a member of a cpio archive for an initrd or initrams is larger than
2Gb, we'll eventually fail to write to that file when we get to that
limit, unless O_LARGEFILE is set.

The problem can be seen with this recipe, assuming that BLK_DEV_RAM
is not configured:

cd /tmp
dd if=/dev/zero of=BIGFILE bs=1048576 count=2200
echo BIGFILE | cpio -o -H newc -R root:root > initrd.img
kexec -l /boot/vmlinuz-$(uname -r) --initrd=initrd.img --reuse-cmdline
kexec -e

The console will show 'Initramfs unpacking failed: write error'.  With
the patch, the error is gone.

Link: https://lkml.kernel.org/r/20240323152934.3307391-1-jsperbeck@google.com
Signed-off-by: John Sperbeck <jsperbeck@google.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
14 months agomm/secretmem: fix GUP-fast succeeding on secretmem folios
David Hildenbrand [Tue, 26 Mar 2024 14:32:08 +0000 (15:32 +0100)]
mm/secretmem: fix GUP-fast succeeding on secretmem folios

folio_is_secretmem() currently relies on secretmem folios being LRU
folios, to save some cycles.

However, folios might reside in a folio batch without the LRU flag set, or
temporarily have their LRU flag cleared.  Consequently, the LRU flag is
unreliable for this purpose.

In particular, this is the case when secretmem_fault() allocates a fresh
page and calls filemap_add_folio()->folio_add_lru().  The folio might be
added to the per-cpu folio batch and won't get the LRU flag set until the
batch was drained using e.g., lru_add_drain().

Consequently, folio_is_secretmem() might not detect secretmem folios and
GUP-fast can succeed in grabbing a secretmem folio, crashing the kernel
when we would later try reading/writing to the folio, because the folio
has been unmapped from the directmap.

Fix it by removing that unreliable check.

Link: https://lkml.kernel.org/r/20240326143210.291116-2-david@redhat.com
Fixes: 1507f51255c9 ("mm: introduce memfd_secret system call to create "secret" memory areas")
Signed-off-by: David Hildenbrand <david@redhat.com>
Reported-by: xingwei lee <xrivendell7@gmail.com>
Reported-by: yue sun <samsun1006219@gmail.com>
Closes: https://lore.kernel.org/lkml/CABOYnLyevJeravW=QrH0JUPYEcDN160aZFb7kwndm-J2rmz0HQ@mail.gmail.com/
Debugged-by: Miklos Szeredi <miklos@szeredi.hu>
Tested-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Mike Rapoport (IBM) <rppt@kernel.org>
Cc: Lorenzo Stoakes <lstoakes@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
14 months agoMerge branch 'acpi-thermal'
Rafael J. Wysocki [Fri, 5 Apr 2024 18:17:48 +0000 (20:17 +0200)]
Merge branch 'acpi-thermal'

* acpi-thermal:
  ACPI: thermal: Register thermal zones without valid trip points

14 months agonfsd: hold a lighter-weight client reference over CB_RECALL_ANY
Jeff Layton [Fri, 5 Apr 2024 17:56:18 +0000 (13:56 -0400)]
nfsd: hold a lighter-weight client reference over CB_RECALL_ANY

Currently the CB_RECALL_ANY job takes a cl_rpc_users reference to the
client. While a callback job is technically an RPC that counter is
really more for client-driven RPCs, and this has the effect of
preventing the client from being unhashed until the callback completes.

If nfsd decides to send a CB_RECALL_ANY just as the client reboots, we
can end up in a situation where the callback can't complete on the (now
dead) callback channel, but the new client can't connect because the old
client can't be unhashed. This usually manifests as a NFS4ERR_DELAY
return on the CREATE_SESSION operation.

The job is only holding a reference to the client so it can clear a flag
after the RPC completes. Fix this by having CB_RECALL_ANY instead hold a
reference to the cl_nfsdfs.cl_ref. Typically we only take that sort of
reference when dealing with the nfsdfs info files, but it should work
appropriately here to ensure that the nfs4_client doesn't disappear.

Fixes: 44df6f439a17 ("NFSD: add delegation reaper to react to low memory condition")
Reported-by: Vladimir Benes <vbenes@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
14 months agoMerge tag '9p-for-6.9-rc3' of https://github.com/martinetd/linux
Linus Torvalds [Fri, 5 Apr 2024 17:05:42 +0000 (10:05 -0700)]
Merge tag '9p-for-6.9-rc3' of https://github.com/martinetd/linux

Pull minor 9p cleanups from Dominique Martinet:

 - kernel doc fix & removal of unused flag

 - fix some bogus debug statement for read/write

* tag '9p-for-6.9-rc3' of https://github.com/martinetd/linux:
  9p: remove SLAB_MEM_SPREAD flag usage
  9p: Fix read/write debug statements to report server reply
  9p/trans_fd: remove Excess kernel-doc comment

14 months agoMerge tag '6.9-rc2-ksmbd-server-fixes' of git://git.samba.org/ksmbd
Linus Torvalds [Fri, 5 Apr 2024 17:02:09 +0000 (10:02 -0700)]
Merge tag '6.9-rc2-ksmbd-server-fixes' of git://git.samba.org/ksmbd

Pull smb server fixes from Steve French:
 "Three fixes, all also for stable:

   - encryption fix

   - memory overrun fix

   - oplock break fix"

* tag '6.9-rc2-ksmbd-server-fixes' of git://git.samba.org/ksmbd:
  ksmbd: do not set SMB2_GLOBAL_CAP_ENCRYPTION for SMB 3.1.1
  ksmbd: validate payload size in ipc response
  ksmbd: don't send oplock break if rename fails

14 months agoMerge tag 'vfs-6.9-rc3.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Linus Torvalds [Fri, 5 Apr 2024 16:47:26 +0000 (09:47 -0700)]
Merge tag 'vfs-6.9-rc3.fixes' of git://git./linux/kernel/git/vfs/vfs

Pull vfs fixes from Christian Brauner:
 "This contains a few small fixes. This comes with some delay because I
  wanted to wait on people running their reproducers and the Easter
  Holidays meant that those replies came in a little later than usual:

   - Fix handling of preventing writes to mounted block devices.

     Since last kernel we allow to prevent writing to mounted block
     devices provided CONFIG_BLK_DEV_WRITE_MOUNTED isn't set and the
     block device is opened with restricted writes. When we switched to
     opening block devices as files we altered the mechanism by which we
     recognize when a block device has been opened with write
     restrictions.

     The detection logic assumed that only read-write mounted
     filesystems would apply write restrictions to their block devices
     from other openers. That of course is not true since it also makes
     sense to apply write restrictions for filesystems that are
     read-only.

     Fix the detection logic using an FMODE_* bit. We still have a few
     left since we freed up a couple a while ago. I also picked up a
     patch to free up four additional FMODE_* bits scheduled for the
     next merge window.

   - Fix counting the number of writers to a block device. This just
     changes the logic to be consistent.

   - Fix a bug in aio causing a NULL pointer derefernce after we
     implemented batched processing in aio.

   - Finally, add the changes we discussed that allows to yield block
     devices early even though file closing itself is deferred.

     This also allows us to remove two holder operations to get and
     release the holder to align lifetime of file and holder of the
     block device"

* tag 'vfs-6.9-rc3.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
  aio: Fix null ptr deref in aio_complete() wakeup
  fs,block: yield devices early
  block: count BLK_OPEN_RESTRICT_WRITES openers
  block: handle BLK_OPEN_RESTRICT_WRITES correctly

14 months agoaio: Fix null ptr deref in aio_complete() wakeup
Kent Overstreet [Sun, 31 Mar 2024 21:52:12 +0000 (17:52 -0400)]
aio: Fix null ptr deref in aio_complete() wakeup

list_del_init_careful() needs to be the last access to the wait queue
entry - it effectively unlocks access.

Previously, finish_wait() would see the empty list head and skip taking
the lock, and then we'd return - but the completion path would still
attempt to do the wakeup after the task_struct pointer had been
overwritten.

Fixes: 71eb6b6b0ba9 ("fs/aio: obey min_nr when doing wakeups")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/linux-fsdevel/CAHTA-ubfwwB51A5Wg5M6H_rPEQK9pNf8FkAGH=vr=FEkyRrtqw@mail.gmail.com/
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Link: https://lore.kernel.org/stable/20240331215212.522544-1-kent.overstreet%40linux.dev
Link: https://lore.kernel.org/r/20240331215212.522544-1-kent.overstreet@linux.dev
Signed-off-by: Christian Brauner <brauner@kernel.org>
14 months agotimers/migration: Return early on deactivation
Anna-Maria Behnsen [Fri, 5 Apr 2024 08:53:21 +0000 (10:53 +0200)]
timers/migration: Return early on deactivation

Commit 4b6f4c5a67c0 ("timer/migration: Remove buggy early return on
deactivation") removed the logic to return early in tmigr_update_events()
on deactivation. With this the problem with a not properly updated first
global event in a hierarchy containing only a single group was fixed.

But when having a look at this code path with a hierarchy with more than a
single level, now unnecessary work is done (example is partially copied
from the message of the commit mentioned above):

                            [GRP1:0]
                         migrator = GRP0:0
                         active   = GRP0:0
                         nextevt  = T0:0i, T0:1
                         /              \
              [GRP0:0]                  [GRP0:1]
           migrator = 0              migrator = NONE
           active   = 0              active   = NONE
           nextevt  = T0i, T1        nextevt  = T2
           /         \                /         \
          0 (T0i)     1 (T1)         2 (T2)      3
      active         idle            idle       idle

0) CPU 0 is active thus its event is ignored (the letter 'i') and so are
upper levels' events. CPU 1 is idle and has the timer T1 enqueued.
CPU 2 also has a timer. The expiry order is T0 (ignored) < T1 < T2

                            [GRP1:0]
                         migrator = GRP0:0
                         active   = GRP0:0
                         nextevt  = T0:0i, T0:1
                         /              \
              [GRP0:0]                  [GRP0:1]
           migrator = NONE           migrator = NONE
           active   = NONE           active   = NONE
           nextevt  = T1             nextevt  = T2
           /         \                /         \
          0 (T0i)     1 (T1)         2 (T2)      3
        idle         idle            idle         idle

1) CPU 0 goes idle without global event queued. Therefore KTIME_MAX is
pushed as its next expiry and its own event kept as "ignore". Without this
early return the following steps happen in tmigr_update_events() when
child = null and group = GRP0:0 :

  lock(GRP0:0->lock);
  timerqueue_del(GRP0:0, T0i);
  unlock(GRP0:0->lock);

                            [GRP1:0]
                         migrator = NONE
                         active   = NONE
                         nextevt  = T0:0, T0:1
                         /              \
              [GRP0:0]                  [GRP0:1]
           migrator = NONE           migrator = NONE
           active   = NONE           active   = NONE
           nextevt  = T1             nextevt  = T2
           /         \                /         \
          0 (T0i)     1 (T1)         2 (T2)      3
        idle         idle            idle         idle

2) The change now propagates up to the top. Then tmigr_update_events()
updates the group event of GRP0:0 and executes the following steps
(child = GRP0:0 and group = GRP0:0):

  lock(GRP0:0->lock);
  lock(GRP1:0->lock);
  evt = tmigr_next_groupevt(GRP0:0); -> this removes the ignored events
in GRP0:0
  ... update GRP1:0 group event and timerqueue ...
  unlock(GRP1:0->lock);
  unlock(GRP0:0->lock);

So the dance in 1) with locking the GRP0:0->lock and removing the T0i from
the timerqueue is redundand as this is done nevertheless in 2) when
tmigr_next_groupevt(GRP0:0) is executed.

Revert commit 4b6f4c5a67c0 ("timer/migration: Remove buggy early return on
deactivation") and add a condition into return path to skip the return
only, when hierarchy contains a single group. Adapt comments accordingly.

Fixes: 4b6f4c5a67c0 ("timer/migration: Remove buggy early return on deactivation")
Signed-off-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Link: https://lore.kernel.org/r/87cyr49on2.fsf@somnus
14 months agotimers/migration: Fix ignored event due to missing CPU update
Frederic Weisbecker [Mon, 1 Apr 2024 21:48:59 +0000 (23:48 +0200)]
timers/migration: Fix ignored event due to missing CPU update

When a group event is updated with its expiry unchanged but a different
CPU, that target change may go unnoticed and the event may be propagated
up with a stale CPU value. The following depicts a scenario that has
been actually observed:

                       [GRP2:0]
                   migrator = GRP1:1
                   active   = GRP1:1
                   nextevt  = TGRP1:0 (T0)
                    /              \
               [GRP1:0]           [GRP1:1]
            migrator = NONE       [...]
            active   = NONE
            nextevt  = TGRP0:0 (T0)
            /           \
        [GRP0:0]       [...]
      migrator = NONE
      active   = NONE
      nextevt  = T0
      /         \
    0 (T0)       1 (T1)
    idle         idle

0) The hierarchy has 3 levels. The left part (GRP1:0) is all idle,
including CPU 0 and CPU 1 which have a timer each: T0 and T1. They have
the same expiry value.

                       [GRP2:0]
                   migrator = GRP1:1
                   active   = GRP1:1
                   nextevt  = KTIME_MAX
                    /              \
               [GRP1:0]           [GRP1:1]
            migrator = NONE       [...]
            active   = NONE
            nextevt  = TGRP0:0 (T0)
            /           \
        [GRP0:0]       [...]
      migrator = NONE
      active   = NONE
      nextevt  = T0
      /         \
    0 (T0)       1 (T1)
    idle         idle

1) The migrator in GRP1:1 handles remotely T0. The event is dequeued
from the top and T0 executed.

                       [GRP2:0]
                   migrator = GRP1:1
                   active   = GRP1:1
                   nextevt  = KTIME_MAX
                    /              \
               [GRP1:0]           [GRP1:1]
            migrator = NONE       [...]
            active   = NONE
            nextevt  = TGRP0:0 (T0)
            /           \
        [GRP0:0]       [...]
      migrator = NONE
      active   = NONE
      nextevt  = T1
      /         \
    0            1 (T1)
    idle         idle

2) The migrator in GRP1:1 fetches the next timer for CPU 0 and finds
none. But it updates the events from its groups, starting with GRP0:0
which now has T1 as its next event. So far so good.

                       [GRP2:0]
                   migrator = GRP1:1
                   active   = GRP1:1
                   nextevt  = KTIME_MAX
                    /              \
               [GRP1:0]           [GRP1:1]
            migrator = NONE       [...]
            active   = NONE
            nextevt  = TGRP0:0 (T0)
            /           \
        [GRP0:0]       [...]
      migrator = NONE
      active   = NONE
      nextevt  = T1
      /         \
    0            1 (T1)
    idle         idle

3) The migrator in GRP1:1 proceeds upward and updates the events in
GRP1:0. The child event TGRP0:0 is found queued with the same expiry
as before. And therefore it is left unchanged. However the target CPU
is not the same but that fact is ignored so TGRP0:0 still points to
CPU 0 when it should point to CPU 1.

                       [GRP2:0]
                   migrator = GRP1:1
                   active   = GRP1:1
                   nextevt  = TGRP1:0 (T0)
                    /              \
               [GRP1:0]           [GRP1:1]
            migrator = NONE       [...]
            active   = NONE
            nextevt  = TGRP0:0 (T0)
            /           \
        [GRP0:0]       [...]
      migrator = NONE
      active   = NONE
      nextevt  = T1
      /         \
    0            1 (T1)
    idle         idle

4) The propagation has reached the top level and TGRP1:0, having TGRP0:0
as its first event, also wrongly points to CPU 0. TGRP1:0 is added to
the top level group.

                       [GRP2:0]
                   migrator = GRP1:1
                   active   = GRP1:1
                   nextevt  = KTIME_MAX
                    /              \
               [GRP1:0]           [GRP1:1]
            migrator = NONE       [...]
            active   = NONE
            nextevt  = TGRP0:0 (T0)
            /           \
        [GRP0:0]       [...]
      migrator = NONE
      active   = NONE
      nextevt  = T1
      /         \
    0            1 (T1)
    idle         idle

5) The migrator in GRP1:1 dequeues the next event in top level pointing
to CPU 0. But since it actually doesn't see any real event in CPU 0, it
early returns.

6) T1 is left unhandled until either CPU 0 or CPU 1 wake up.

Some other bad scenario may involve trees with just two levels.

Fix this with unconditionally updating the CPU of the child event before
considering to early return while updating a queued event with an
unchanged expiry value.

Fixes: 7ee988770326 ("timers: Implement the hierarchical pull model")
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Link: https://lore.kernel.org/r/Zg2Ct6M2RJAYHgCB@localhost.localdomain
14 months agoMerge tag 'asoc-fix-v6.9-rc2' of https://git.kernel.org/pub/scm/linux/kernel/git...
Takashi Iwai [Fri, 5 Apr 2024 06:48:12 +0000 (08:48 +0200)]
Merge tag 'asoc-fix-v6.9-rc2' of https://git./linux/kernel/git/broonie/sound into for-linus

ASoC: Fixes for v6.9

A relatively large set of fixes here, the biggest piece of it is a
series correcting some problems with the delay reporting for Intel SOF
cards but there's a bunch of other things.  Everything here is driver
specific except for a fix in the core for an issue with sign extension
handling volume controls.

14 months agoMerge tag 'drm-intel-fixes-2024-04-04' of https://anongit.freedesktop.org/git/drm...
Dave Airlie [Fri, 5 Apr 2024 02:32:08 +0000 (12:32 +1000)]
Merge tag 'drm-intel-fixes-2024-04-04' of https://anongit.freedesktop.org/git/drm/drm-intel into drm-fixes

Display fixes:
- A few DisplayPort related fixes (Imre, Arun, Ankit, Ville)
- eDP PSR fixes (Jouni)

Core/GT fixes:
- Remove some VM space restrictions on older platforms (Andi)
- Disable automatic load CCS load balancing (Andi)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/Zg7nSK5oTmWfKPPI@intel.com
14 months agoMerge tag 'drm-xe-fixes-2024-04-04' of https://gitlab.freedesktop.org/drm/xe/kernel...
Dave Airlie [Fri, 5 Apr 2024 02:25:28 +0000 (12:25 +1000)]
Merge tag 'drm-xe-fixes-2024-04-04' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-fixes

- Stop using system_unbound_wq for preempt fences,
  as this can cause starvation when reaching more
  than max_active defined by workqueue
- Fix saving unordered rebinding fences by attaching
  them as kernel feces to the vm's resv
- Fix TLB invalidation fences completing out of order
- Move rebind TLB invalidation to the ring ops to reduce
  the latency

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/tizan6wdpxu4ayudeikjglxdgzmnhdzj3li3z2pgkierjtozzw@lbfddeg43a7h
14 months agoMerge tag 'drm-misc-fixes-2024-04-04' of https://gitlab.freedesktop.org/drm/misc...
Dave Airlie [Fri, 5 Apr 2024 01:59:02 +0000 (11:59 +1000)]
Merge tag 'drm-misc-fixes-2024-04-04' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-fixes

Short summary of fixes pull:

display:
- fix typos in kerneldoc

nouveau:
- uvmm: fix remap address calculation
- minor cleanups

panfrost:
- fix power-transition timeouts

prime:
- unbreak dma-buf export for virt-gpu

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20240404104813.GA27376@localhost.localdomain
14 months agox86/cpufeatures: Add CPUID_LNX_5 to track recently added Linux-defined word
Sean Christopherson [Fri, 5 Apr 2024 00:16:14 +0000 (17:16 -0700)]
x86/cpufeatures: Add CPUID_LNX_5 to track recently added Linux-defined word

Add CPUID_LNX_5 to track cpufeatures' word 21, and add the appropriate
compile-time assert in KVM to prevent direct lookups on the features in
CPUID_LNX_5.  KVM uses X86_FEATURE_* flags to manage guest CPUID, and so
must translate features that are scattered by Linux from the Linux-defined
bit to the hardware-defined bit, i.e. should never try to directly access
scattered features in guest CPUID.

Opportunistically add NR_CPUID_WORDS to enum cpuid_leafs, along with a
compile-time assert in KVM's CPUID infrastructure to ensure that future
additions update cpuid_leafs along with NCAPINTS.

No functional change intended.

Fixes: 7f274e609f3d ("x86/cpufeatures: Add new word for scattered features")
Cc: Sandipan Das <sandipan.das@amd.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Acked-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 months agoMerge tag 'net-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 4 Apr 2024 21:49:10 +0000 (14:49 -0700)]
Merge tag 'net-6.9-rc3' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from netfilter, bluetooth and bpf.

  Fairly usual collection of driver and core fixes. The large selftest
  accompanying one of the fixes is also becoming a common occurrence.

  Current release - regressions:

   - ipv6: fix infinite recursion in fib6_dump_done()

   - net/rds: fix possible null-deref in newly added error path

  Current release - new code bugs:

   - net: do not consume a full cacheline for system_page_pool

   - bpf: fix bpf_arena-related file descriptor leaks in the verifier

   - drv: ice: fix freeing uninitialized pointers, fixing misuse of the
     newfangled __free() auto-cleanup

  Previous releases - regressions:

   - x86/bpf: fixes the BPF JIT with retbleed=stuff

   - xen-netfront: add missing skb_mark_for_recycle, fix page pool
     accounting leaks, revealed by recently added explicit warning

   - tcp: fix bind() regression for v6-only wildcard and v4-mapped-v6
     non-wildcard addresses

   - Bluetooth:
      - replace "hci_qca: Set BDA quirk bit if fwnode exists in DT" with
        better workarounds to un-break some buggy Qualcomm devices
      - set conn encrypted before conn establishes, fix re-connecting to
        some headsets which use slightly unusual sequence of msgs

   - mptcp:
      - prevent BPF accessing lowat from a subflow socket
      - don't account accept() of non-MPC client as fallback to TCP

   - drv: mana: fix Rx DMA datasize and skb_over_panic

   - drv: i40e: fix VF MAC filter removal

  Previous releases - always broken:

   - gro: various fixes related to UDP tunnels - netns crossing
     problems, incorrect checksum conversions, and incorrect packet
     transformations which may lead to panics

   - bpf: support deferring bpf_link dealloc to after RCU grace period

   - nf_tables:
      - release batch on table validation from abort path
      - release mutex after nft_gc_seq_end from abort path
      - flush pending destroy work before exit_net release

   - drv: r8169: skip DASH fw status checks when DASH is disabled"

* tag 'net-6.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (81 commits)
  netfilter: validate user input for expected length
  net/sched: act_skbmod: prevent kernel-infoleak
  net: usb: ax88179_178a: avoid the interface always configured as random address
  net: dsa: sja1105: Fix parameters order in sja1110_pcs_mdio_write_c45()
  net: ravb: Always update error counters
  net: ravb: Always process TX descriptor ring
  netfilter: nf_tables: discard table flag update with pending basechain deletion
  netfilter: nf_tables: Fix potential data-race in __nft_flowtable_type_get()
  netfilter: nf_tables: reject new basechain after table flag update
  netfilter: nf_tables: flush pending destroy work before exit_net release
  netfilter: nf_tables: release mutex after nft_gc_seq_end from abort path
  netfilter: nf_tables: release batch on table validation from abort path
  Revert "tg3: Remove residual error handling in tg3_suspend"
  tg3: Remove residual error handling in tg3_suspend
  net: mana: Fix Rx DMA datasize and skb_over_panic
  net/sched: fix lockdep splat in qdisc_tree_reduce_backlog()
  net: phy: micrel: lan8814: Fix when enabling/disabling 1-step timestamping
  net: stmmac: fix rx queue priority assignment
  net: txgbe: fix i2c dev name cannot match clkdev
  net: fec: Set mac_managed_pm during probe
  ...

14 months agoMerge tag 'bcachefs-2024-04-03' of https://evilpiepirate.org/git/bcachefs
Linus Torvalds [Thu, 4 Apr 2024 21:36:32 +0000 (14:36 -0700)]
Merge tag 'bcachefs-2024-04-03' of https://evilpiepirate.org/git/bcachefs

Pull bcachefs repair code from Kent Overstreet:
 "A couple more small fixes, and new repair code.

  We can now automatically recover from arbitrary corrupted interior
  btree nodes by scanning, and we can reconstruct metadata as needed to
  bring a filesystem back into a working, consistent, read-write state
  and preserve access to whatevver wasn't corrupted.

  Meaning - you can blow away all metadata except for extents and
  dirents leaf nodes, and repair will reconstruct everything else and
  give you your data, and under the correct paths. If inodes are missing
  i_size will be slightly off and permissions/ownership/timestamps will
  be gone, and we do still need the snapshots btree if snapshots were in
  use - in the future we'll be able to guess the snapshot tree structure
  in some situations.

  IOW - aside from shaking out remaining bugs (fuzz testing is still
  coming), repair code should be complete and if repair ever doesn't
  work that's the highest priority bug that I want to know about
  immediately.

  This patchset was kindly tested by a user from India who accidentally
  wiped one drive out of a three drive filesystem with no replication on
  the family computer - it took a couple weeks but we got everything
  important back"

* tag 'bcachefs-2024-04-03' of https://evilpiepirate.org/git/bcachefs:
  bcachefs: reconstruct_inode()
  bcachefs: Subvolume reconstruction
  bcachefs: Check for extents that point to same space
  bcachefs: Reconstruct missing snapshot nodes
  bcachefs: Flag btrees with missing data
  bcachefs: Topology repair now uses nodes found by scanning to fill holes
  bcachefs: Repair pass for scanning for btree nodes
  bcachefs: Don't skip fake btree roots in fsck
  bcachefs: bch2_btree_root_alloc() -> bch2_btree_root_alloc_fake()
  bcachefs: Etyzinger cleanups
  bcachefs: bch2_shoot_down_journal_keys()
  bcachefs: Clear recovery_passes_required as they complete without errors
  bcachefs: ratelimit informational fsck errors
  bcachefs: Check for bad needs_discard before doing discard
  bcachefs: Improve bch2_btree_update_to_text()
  mean_and_variance: Drop always failing tests
  bcachefs: fix nocow lock deadlock
  bcachefs: BCH_WATERMARK_interior_updates
  bcachefs: Fix btree node reserve

14 months agoriscv: process: Fix kernel gp leakage
Stefan O'Rear [Wed, 27 Mar 2024 06:12:58 +0000 (02:12 -0400)]
riscv: process: Fix kernel gp leakage

childregs represents the registers which are active for the new thread
in user context. For a kernel thread, childregs->gp is never used since
the kernel gp is not touched by switch_to. For a user mode helper, the
gp value can be observed in user space after execve or possibly by other
means.

[From the email thread]

The /* Kernel thread */ comment is somewhat inaccurate in that it is also used
for user_mode_helper threads, which exec a user process, e.g. /sbin/init or
when /proc/sys/kernel/core_pattern is a pipe. Such threads do not have
PF_KTHREAD set and are valid targets for ptrace etc. even before they exec.

childregs is the *user* context during syscall execution and it is observable
from userspace in at least five ways:

1. kernel_execve does not currently clear integer registers, so the starting
   register state for PID 1 and other user processes started by the kernel has
   sp = user stack, gp = kernel __global_pointer$, all other integer registers
   zeroed by the memset in the patch comment.

   This is a bug in its own right, but I'm unwilling to bet that it is the only
   way to exploit the issue addressed by this patch.

2. ptrace(PTRACE_GETREGSET): you can PTRACE_ATTACH to a user_mode_helper thread
   before it execs, but ptrace requires SIGSTOP to be delivered which can only
   happen at user/kernel boundaries.

3. /proc/*/task/*/syscall: this is perfectly happy to read pt_regs for
   user_mode_helpers before the exec completes, but gp is not one of the
   registers it returns.

4. PERF_SAMPLE_REGS_USER: LOCKDOWN_PERF normally prevents access to kernel
   addresses via PERF_SAMPLE_REGS_INTR, but due to this bug kernel addresses
   are also exposed via PERF_SAMPLE_REGS_USER which is permitted under
   LOCKDOWN_PERF. I have not attempted to write exploit code.

5. Much of the tracing infrastructure allows access to user registers. I have
   not attempted to determine which forms of tracing allow access to user
   registers without already allowing access to kernel registers.

Fixes: 7db91e57a0ac ("RISC-V: Task implementation")
Cc: stable@vger.kernel.org
Signed-off-by: Stefan O'Rear <sorear@fastmail.com>
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20240327061258.2370291-1-sorear@fastmail.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
14 months agoriscv: Disable preemption when using patch_map()
Alexandre Ghiti [Tue, 26 Mar 2024 20:30:17 +0000 (21:30 +0100)]
riscv: Disable preemption when using patch_map()

patch_map() uses fixmap mappings to circumvent the non-writability of
the kernel text mapping.

The __set_fixmap() function only flushes the current cpu tlb, it does
not emit an IPI so we must make sure that while we use a fixmap mapping,
the current task is not migrated on another cpu which could miss the
newly introduced fixmap mapping.

So in order to avoid any task migration, disable the preemption.

Reported-by: Andrea Parri <andrea@rivosinc.com>
Closes: https://lore.kernel.org/all/ZcS+GAaM25LXsBOl@andrea/
Reported-by: Andy Chiu <andy.chiu@sifive.com>
Closes: https://lore.kernel.org/linux-riscv/CABgGipUMz3Sffu-CkmeUB1dKVwVQ73+7=sgC45-m0AE9RCjOZg@mail.gmail.com/
Fixes: cad539baa48f ("riscv: implement a memset like function for text")
Fixes: 0ff7c3b33127 ("riscv: Use text_mutex instead of patch_lock")
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Signed-off-by: Andy Chiu <andy.chiu@sifive.com>
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Acked-by: Puranjay Mohan <puranjay12@gmail.com>
Link: https://lore.kernel.org/r/20240326203017.310422-3-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
14 months agoriscv: Fix warning by declaring arch_cpu_idle() as noinstr
Alexandre Ghiti [Tue, 26 Mar 2024 20:30:16 +0000 (21:30 +0100)]
riscv: Fix warning by declaring arch_cpu_idle() as noinstr

The following warning appears when using ftrace:

[89855.443413] RCU not on for: arch_cpu_idle+0x0/0x1c
[89855.445640] WARNING: CPU: 5 PID: 0 at include/linux/trace_recursion.h:162 arch_ftrace_ops_list_func+0x208/0x228
[89855.445824] Modules linked in: xt_conntrack(E) nft_chain_nat(E) xt_MASQUERADE(E) nf_conntrack_netlink(E) xt_addrtype(E) nft_compat(E) nf_tables(E) nfnetlink(E) br_netfilter(E) cfg80211(E) nls_iso8859_1(E) ofpart(E) redboot(E) cmdlinepart(E) cfi_cmdset_0001(E) virtio_net(E) cfi_probe(E) cfi_util(E) 9pnet_virtio(E) gen_probe(E) net_failover(E) virtio_rng(E) failover(E) 9pnet(E) physmap(E) map_funcs(E) chipreg(E) mtd(E) uio_pdrv_genirq(E) uio(E) dm_multipath(E) scsi_dh_rdac(E) scsi_dh_emc(E) scsi_dh_alua(E) drm(E) efi_pstore(E) backlight(E) ip_tables(E) x_tables(E) raid10(E) raid456(E) async_raid6_recov(E) async_memcpy(E) async_pq(E) async_xor(E) xor(E) async_tx(E) raid6_pq(E) raid1(E) raid0(E) virtio_blk(E)
[89855.451563] CPU: 5 PID: 0 Comm: swapper/5 Tainted: G            E      6.8.0-rc6ubuntu-defconfig #2
[89855.451726] Hardware name: riscv-virtio,qemu (DT)
[89855.451899] epc : arch_ftrace_ops_list_func+0x208/0x228
[89855.452016]  ra : arch_ftrace_ops_list_func+0x208/0x228
[89855.452119] epc : ffffffff8016b216 ra : ffffffff8016b216 sp : ffffaf808090fdb0
[89855.452171]  gp : ffffffff827c7680 tp : ffffaf808089ad40 t0 : ffffffff800c0dd8
[89855.452216]  t1 : 0000000000000001 t2 : 0000000000000000 s0 : ffffaf808090fe30
[89855.452306]  s1 : 0000000000000000 a0 : 0000000000000026 a1 : ffffffff82cd6ac8
[89855.452423]  a2 : ffffffff800458c8 a3 : ffffaf80b1870640 a4 : 0000000000000000
[89855.452646]  a5 : 0000000000000000 a6 : 00000000ffffffff a7 : ffffffffffffffff
[89855.452698]  s2 : ffffffff82766872 s3 : ffffffff80004caa s4 : ffffffff80ebea90
[89855.452743]  s5 : ffffaf808089bd40 s6 : 8000000a00006e00 s7 : 0000000000000008
[89855.452787]  s8 : 0000000000002000 s9 : 0000000080043700 s10: 0000000000000000
[89855.452831]  s11: 0000000000000000 t3 : 0000000000100000 t4 : 0000000000000064
[89855.452874]  t5 : 000000000000000c t6 : ffffaf80b182dbfc
[89855.452929] status: 0000000200000100 badaddr: 0000000000000000 cause: 0000000000000003
[89855.453053] [<ffffffff8016b216>] arch_ftrace_ops_list_func+0x208/0x228
[89855.453191] [<ffffffff8000e082>] ftrace_call+0x8/0x22
[89855.453265] [<ffffffff800a149c>] do_idle+0x24c/0x2ca
[89855.453357] [<ffffffff8000da54>] return_to_handler+0x0/0x26
[89855.453429] [<ffffffff8000b716>] smp_callin+0x92/0xb6
[89855.453785] ---[ end trace 0000000000000000 ]---

To fix this, mark arch_cpu_idle() as noinstr, like it is done in commit
a9cbc1b471d2 ("s390/idle: mark arch_cpu_idle() noinstr").

Reported-by: Evgenii Shatokhin <e.shatokhin@yadro.com>
Closes: https://lore.kernel.org/linux-riscv/51f21b87-ebed-4411-afbc-c00d3dea2bab@yadro.com/
Fixes: cfbc4f81c9d0 ("riscv: Select ARCH_WANTS_NO_INSTR")
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Reviewed-by: Andy Chiu <andy.chiu@sifive.com>
Tested-by: Andy Chiu <andy.chiu@sifive.com>
Acked-by: Puranjay Mohan <puranjay12@gmail.com>
Link: https://lore.kernel.org/r/20240326203017.310422-2-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
14 months agoMerge tag 'nvme-6.9-2024-04-04' of git://git.infradead.org/nvme into block-6.9
Jens Axboe [Thu, 4 Apr 2024 19:23:21 +0000 (13:23 -0600)]
Merge tag 'nvme-6.9-2024-04-04' of git://git.infradead.org/nvme into block-6.9

Pull NVMe fixes from Keith:

"nvme fixes for Linux 6.9

 - Atomic queue limits fixes (Christoph)
 - Fabrics fixes (Hannes, Daniel)"

* tag 'nvme-6.9-2024-04-04' of git://git.infradead.org/nvme:
  nvme-fc: rename free_ctrl callback to match name pattern
  nvmet-fc: move RCU read lock to nvmet_fc_assoc_exists
  nvmet: implement unique discovery NQN
  nvme: don't create a multipath node for zero capacity devices
  nvme: split nvme_update_zone_info
  nvme-multipath: don't inherit LBA-related fields for the multipath node

14 months agoriscv: use KERN_INFO in do_trap
Andreas Schwab [Tue, 19 Mar 2024 15:40:05 +0000 (16:40 +0100)]
riscv: use KERN_INFO in do_trap

Print the instruction dump with info instead of emergency level.  The
unhandled signal message is only for informational purpose.

Fixes: b8a03a634129 ("riscv: add userland instruction dump to RISC-V splats")
Signed-off-by: Andreas Schwab <schwab@suse.de>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Reviewed-by: Yunhui Cui <cuiyunhui@bytedance.com>
Link: https://lore.kernel.org/r/mvmy1aegrhm.fsf@suse.de
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
14 months agoASoC: SOF: Core: Add remove_late() to sof_init_environment failure path
Chaitanya Kumar Borah [Thu, 4 Apr 2024 18:48:13 +0000 (13:48 -0500)]
ASoC: SOF: Core: Add remove_late() to sof_init_environment failure path

In cases where the sof driver is unable to find the firmware and/or
topology file [1], it exits without releasing the i915 runtime
pm wakeref [2]. This results in dmesg warnings[3] during
suspend/resume or driver unbind. Add remove_late() to the failure path
of sof_init_environment so that i915 wakeref is released appropriately

[1]

[    8.990366] sof-audio-pci-intel-mtl 0000:00:1f.3: SOF firmware and/or topology file not found.
[    8.990396] sof-audio-pci-intel-mtl 0000:00:1f.3: Supported default profiles
[    8.990398] sof-audio-pci-intel-mtl 0000:00:1f.3: - ipc type 1 (Requested):
[    8.990399] sof-audio-pci-intel-mtl 0000:00:1f.3:  Firmware file: intel/sof-ipc4/mtl/sof-mtl.ri
[    8.990401] sof-audio-pci-intel-mtl 0000:00:1f.3:  Topology file: intel/sof-ace-tplg/sof-mtl-rt711-2ch.tplg
[    8.990402] sof-audio-pci-intel-mtl 0000:00:1f.3: Check if you have 'sof-firmware' package installed.
[    8.990403] sof-audio-pci-intel-mtl 0000:00:1f.3: Optionally it can be manually downloaded from:
[    8.990404] sof-audio-pci-intel-mtl 0000:00:1f.3:    https://github.com/thesofproject/sof-bin/
[    8.999088] sof-audio-pci-intel-mtl 0000:00:1f.3: error: sof_probe_work failed err: -2

[2]

ref_tracker: 0000:00:02.0@ffff9b8511b6a378 has 1/5 users at
     track_intel_runtime_pm_wakeref.part.0+0x36/0x70 [i915]
     __intel_runtime_pm_get+0x51/0xb0 [i915]
     intel_runtime_pm_get+0x17/0x20 [i915]
     intel_display_power_get+0x2f/0x70 [i915]
     i915_audio_component_get_power+0x23/0x120 [i915]
     snd_hdac_display_power+0x89/0x130 [snd_hda_core]
     hda_codec_i915_init+0x3f/0x50 [snd_sof_intel_hda]
     hda_dsp_probe_early+0x170/0x250 [snd_sof_intel_hda_common]
     snd_sof_device_probe+0x224/0x320 [snd_sof]
     sof_pci_probe+0x15b/0x220 [snd_sof_pci]
     hda_pci_intel_probe+0x30/0x70 [snd_sof_intel_hda_common]
     local_pci_probe+0x4c/0xb0
     pci_device_probe+0xcc/0x250
     really_probe+0x18e/0x420
     __driver_probe_device+0x7e/0x170
     driver_probe_device+0x23/0xa0

[3]
[  484.105070] ------------[ cut here ]------------
[  484.108238] thunderbolt 0000:00:0d.2: PM: pci_pm_suspend_late+0x0/0x50 returned 0 after 0 usecs
[  484.117106] i915 0000:00:02.0: i915 raw-wakerefs=1 wakelocks=1 on cleanup
[  484.792005] WARNING: CPU: 2 PID: 2405 at drivers/gpu/drm/i915/intel_runtime_pm.c:444 intel_runtime_pm_driver_release+0x6c/0x80

Tested-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Reviewed-by: Péter Ujfalusi <peter.ujfalusi@linux.intel.com>
Reviewed-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Signed-off-by: Chaitanya Kumar Borah <chaitanya.kumar.borah@intel.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://github.com/thesofproject/linux/pull/4878
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://msgid.link/r/20240404184813.134566-1-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
14 months agoMerge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Jakub Kicinski [Thu, 4 Apr 2024 18:37:39 +0000 (11:37 -0700)]
Merge tag 'for-netdev' of https://git./linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2024-04-04

We've added 7 non-merge commits during the last 5 day(s) which contain
a total of 9 files changed, 75 insertions(+), 24 deletions(-).

The main changes are:

1) Fix x86 BPF JIT under retbleed=stuff which causes kernel panics due to
   incorrect destination IP calculation and incorrect IP for relocations,
   from Uros Bizjak and Joan Bruguera Micó.

2) Fix BPF arena file descriptor leaks in the verifier,
   from Anton Protopopov.

3) Defer bpf_link deallocation to after RCU grace period as currently
   running multi-{kprobes,uprobes} programs might still access cookie
   information from the link, from Andrii Nakryiko.

4) Fix a BPF sockmap lock inversion deadlock in map_delete_elem reported
   by syzkaller, from Jakub Sitnicki.

5) Fix resolve_btfids build with musl libc due to missing linux/types.h
   include, from Natanael Copa.

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
  bpf, sockmap: Prevent lock inversion deadlock in map delete elem
  x86/bpf: Fix IP for relocating call depth accounting
  x86/bpf: Fix IP after emitting call depth accounting
  bpf: fix possible file descriptor leaks in verifier
  tools/resolve_btfids: fix build with musl libc
  bpf: support deferring bpf_link dealloc to after RCU grace period
  bpf: put uprobe link's path and task in release callback
====================

Link: https://lore.kernel.org/r/20240404183258.4401-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
14 months agoPM: EM: fix wrong utilization estimation in em_cpu_energy()
Vincent Guittot [Thu, 4 Apr 2024 10:42:00 +0000 (12:42 +0200)]
PM: EM: fix wrong utilization estimation in em_cpu_energy()

Commit 1b600da51073 ("PM: EM: Optimize em_cpu_energy() and remove division")
has added back map_util_perf() in em_cpu_energy() computation which has
been removed with the rework of scheduler/cpufreq interface.
This is wrong because sugov_effective_cpu_perf() already takes care of
mapping the utilization to a performance level.

Fixes: 1b600da51073 ("PM: EM: Optimize em_cpu_energy() and remove division")
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
14 months agogpio: cdev: fix missed label sanitizing in debounce_setup()
Kent Gibson [Thu, 4 Apr 2024 09:33:28 +0000 (11:33 +0200)]
gpio: cdev: fix missed label sanitizing in debounce_setup()

When adding sanitization of the label, the path through
edge_detector_setup() that leads to debounce_setup() was overlooked.
A request taking this path does not allocate a new label and the
request label is freed twice when the request is released, resulting
in memory corruption.

Add label sanitization to debounce_setup().

Cc: stable@vger.kernel.org
Fixes: b34490879baa ("gpio: cdev: sanitize the label before requesting the interrupt")
Signed-off-by: Kent Gibson <warthog618@gmail.com>
[Bartosz: rebased on top of the fix for empty GPIO labels]
Co-developed-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
14 months agonetfilter: validate user input for expected length
Eric Dumazet [Thu, 4 Apr 2024 12:20:51 +0000 (12:20 +0000)]
netfilter: validate user input for expected length

I got multiple syzbot reports showing old bugs exposed
by BPF after commit 20f2505fb436 ("bpf: Try to avoid kzalloc
in cgroup/{s,g}etsockopt")

setsockopt() @optlen argument should be taken into account
before copying data.

 BUG: KASAN: slab-out-of-bounds in copy_from_sockptr_offset include/linux/sockptr.h:49 [inline]
 BUG: KASAN: slab-out-of-bounds in copy_from_sockptr include/linux/sockptr.h:55 [inline]
 BUG: KASAN: slab-out-of-bounds in do_replace net/ipv4/netfilter/ip_tables.c:1111 [inline]
 BUG: KASAN: slab-out-of-bounds in do_ipt_set_ctl+0x902/0x3dd0 net/ipv4/netfilter/ip_tables.c:1627
Read of size 96 at addr ffff88802cd73da0 by task syz-executor.4/7238

CPU: 1 PID: 7238 Comm: syz-executor.4 Not tainted 6.9.0-rc2-next-20240403-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 03/27/2024
Call Trace:
 <TASK>
  __dump_stack lib/dump_stack.c:88 [inline]
  dump_stack_lvl+0x241/0x360 lib/dump_stack.c:114
  print_address_description mm/kasan/report.c:377 [inline]
  print_report+0x169/0x550 mm/kasan/report.c:488
  kasan_report+0x143/0x180 mm/kasan/report.c:601
  kasan_check_range+0x282/0x290 mm/kasan/generic.c:189
  __asan_memcpy+0x29/0x70 mm/kasan/shadow.c:105
  copy_from_sockptr_offset include/linux/sockptr.h:49 [inline]
  copy_from_sockptr include/linux/sockptr.h:55 [inline]
  do_replace net/ipv4/netfilter/ip_tables.c:1111 [inline]
  do_ipt_set_ctl+0x902/0x3dd0 net/ipv4/netfilter/ip_tables.c:1627
  nf_setsockopt+0x295/0x2c0 net/netfilter/nf_sockopt.c:101
  do_sock_setsockopt+0x3af/0x720 net/socket.c:2311
  __sys_setsockopt+0x1ae/0x250 net/socket.c:2334
  __do_sys_setsockopt net/socket.c:2343 [inline]
  __se_sys_setsockopt net/socket.c:2340 [inline]
  __x64_sys_setsockopt+0xb5/0xd0 net/socket.c:2340
 do_syscall_64+0xfb/0x240
 entry_SYSCALL_64_after_hwframe+0x72/0x7a
RIP: 0033:0x7fd22067dde9
Code: 28 00 00 00 75 05 48 83 c4 28 c3 e8 e1 20 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 b0 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007fd21f9ff0c8 EFLAGS: 00000246 ORIG_RAX: 0000000000000036
RAX: ffffffffffffffda RBX: 00007fd2207abf80 RCX: 00007fd22067dde9
RDX: 0000000000000040 RSI: 0000000000000000 RDI: 0000000000000003
RBP: 00007fd2206ca47a R08: 0000000000000001 R09: 0000000000000000
R10: 0000000020000880 R11: 0000000000000246 R12: 0000000000000000
R13: 000000000000000b R14: 00007fd2207abf80 R15: 00007ffd2d0170d8
 </TASK>

Allocated by task 7238:
  kasan_save_stack mm/kasan/common.c:47 [inline]
  kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
  poison_kmalloc_redzone mm/kasan/common.c:370 [inline]
  __kasan_kmalloc+0x98/0xb0 mm/kasan/common.c:387
  kasan_kmalloc include/linux/kasan.h:211 [inline]
  __do_kmalloc_node mm/slub.c:4069 [inline]
  __kmalloc_noprof+0x200/0x410 mm/slub.c:4082
  kmalloc_noprof include/linux/slab.h:664 [inline]
  __cgroup_bpf_run_filter_setsockopt+0xd47/0x1050 kernel/bpf/cgroup.c:1869
  do_sock_setsockopt+0x6b4/0x720 net/socket.c:2293
  __sys_setsockopt+0x1ae/0x250 net/socket.c:2334
  __do_sys_setsockopt net/socket.c:2343 [inline]
  __se_sys_setsockopt net/socket.c:2340 [inline]
  __x64_sys_setsockopt+0xb5/0xd0 net/socket.c:2340
 do_syscall_64+0xfb/0x240
 entry_SYSCALL_64_after_hwframe+0x72/0x7a

The buggy address belongs to the object at ffff88802cd73da0
 which belongs to the cache kmalloc-8 of size 8
The buggy address is located 0 bytes inside of
 allocated 1-byte region [ffff88802cd73da0ffff88802cd73da1)

The buggy address belongs to the physical page:
page: refcount:1 mapcount:0 mapping:0000000000000000 index:0xffff88802cd73020 pfn:0x2cd73
flags: 0xfff80000000000(node=0|zone=1|lastcpupid=0xfff)
page_type: 0xffffefff(slab)
raw: 00fff80000000000 ffff888015041280 dead000000000100 dead000000000122
raw: ffff88802cd73020 000000008080007f 00000001ffffefff 0000000000000000
page dumped because: kasan: bad access detected
page_owner tracks the page as allocated
page last allocated via order 0, migratetype Unmovable, gfp_mask 0x12cc0(GFP_KERNEL|__GFP_NOWARN|__GFP_NORETRY), pid 5103, tgid 2119833701 (syz-executor.4), ts 5103, free_ts 70804600828
  set_page_owner include/linux/page_owner.h:32 [inline]
  post_alloc_hook+0x1f3/0x230 mm/page_alloc.c:1490
  prep_new_page mm/page_alloc.c:1498 [inline]
  get_page_from_freelist+0x2e7e/0x2f40 mm/page_alloc.c:3454
  __alloc_pages_noprof+0x256/0x6c0 mm/page_alloc.c:4712
  __alloc_pages_node_noprof include/linux/gfp.h:244 [inline]
  alloc_pages_node_noprof include/linux/gfp.h:271 [inline]
  alloc_slab_page+0x5f/0x120 mm/slub.c:2249
  allocate_slab+0x5a/0x2e0 mm/slub.c:2412
  new_slab mm/slub.c:2465 [inline]
  ___slab_alloc+0xcd1/0x14b0 mm/slub.c:3615
  __slab_alloc+0x58/0xa0 mm/slub.c:3705
  __slab_alloc_node mm/slub.c:3758 [inline]
  slab_alloc_node mm/slub.c:3936 [inline]
  __do_kmalloc_node mm/slub.c:4068 [inline]
  kmalloc_node_track_caller_noprof+0x286/0x450 mm/slub.c:4089
  kstrdup+0x3a/0x80 mm/util.c:62
  device_rename+0xb5/0x1b0 drivers/base/core.c:4558
  dev_change_name+0x275/0x860 net/core/dev.c:1232
  do_setlink+0xa4b/0x41f0 net/core/rtnetlink.c:2864
  __rtnl_newlink net/core/rtnetlink.c:3680 [inline]
  rtnl_newlink+0x180b/0x20a0 net/core/rtnetlink.c:3727
  rtnetlink_rcv_msg+0x89b/0x10d0 net/core/rtnetlink.c:6594
  netlink_rcv_skb+0x1e3/0x430 net/netlink/af_netlink.c:2559
  netlink_unicast_kernel net/netlink/af_netlink.c:1335 [inline]
  netlink_unicast+0x7ea/0x980 net/netlink/af_netlink.c:1361
page last free pid 5146 tgid 5146 stack trace:
  reset_page_owner include/linux/page_owner.h:25 [inline]
  free_pages_prepare mm/page_alloc.c:1110 [inline]
  free_unref_page+0xd3c/0xec0 mm/page_alloc.c:2617
  discard_slab mm/slub.c:2511 [inline]
  __put_partials+0xeb/0x130 mm/slub.c:2980
  put_cpu_partial+0x17c/0x250 mm/slub.c:3055
  __slab_free+0x2ea/0x3d0 mm/slub.c:4254
  qlink_free mm/kasan/quarantine.c:163 [inline]
  qlist_free_all+0x9e/0x140 mm/kasan/quarantine.c:179
  kasan_quarantine_reduce+0x14f/0x170 mm/kasan/quarantine.c:286
  __kasan_slab_alloc+0x23/0x80 mm/kasan/common.c:322
  kasan_slab_alloc include/linux/kasan.h:201 [inline]
  slab_post_alloc_hook mm/slub.c:3888 [inline]
  slab_alloc_node mm/slub.c:3948 [inline]
  __do_kmalloc_node mm/slub.c:4068 [inline]
  __kmalloc_node_noprof+0x1d7/0x450 mm/slub.c:4076
  kmalloc_node_noprof include/linux/slab.h:681 [inline]
  kvmalloc_node_noprof+0x72/0x190 mm/util.c:634
  bucket_table_alloc lib/rhashtable.c:186 [inline]
  rhashtable_rehash_alloc+0x9e/0x290 lib/rhashtable.c:367
  rht_deferred_worker+0x4e1/0x2440 lib/rhashtable.c:427
  process_one_work kernel/workqueue.c:3218 [inline]
  process_scheduled_works+0xa2c/0x1830 kernel/workqueue.c:3299
  worker_thread+0x86d/0xd70 kernel/workqueue.c:3380
  kthread+0x2f0/0x390 kernel/kthread.c:388
  ret_from_fork+0x4b/0x80 arch/x86/kernel/process.c:147
  ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:243

Memory state around the buggy address:
 ffff88802cd73c80: 07 fc fc fc 05 fc fc fc 05 fc fc fc fa fc fc fc
 ffff88802cd73d00: fa fc fc fc fa fc fc fc fa fc fc fc fa fc fc fc
>ffff88802cd73d80: fa fc fc fc 01 fc fc fc fa fc fc fc fa fc fc fc
                               ^
 ffff88802cd73e00: fa fc fc fc fa fc fc fc 05 fc fc fc 07 fc fc fc
 ffff88802cd73e80: 07 fc fc fc 07 fc fc fc 07 fc fc fc 07 fc fc fc

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Pablo Neira Ayuso <pablo@netfilter.org>
Link: https://lore.kernel.org/r/20240404122051.2303764-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
14 months agoMerge tag 'nf-24-04-04' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Jakub Kicinski [Thu, 4 Apr 2024 16:38:52 +0000 (09:38 -0700)]
Merge tag 'nf-24-04-04' of git://git./linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contains Netfilter fixes for net:

Patch #1 unlike early commit path stage which triggers a call to abort,
         an explicit release of the batch is required on abort, otherwise
         mutex is released and commit_list remains in place.

Patch #2 release mutex after nft_gc_seq_end() in commit path, otherwise
         async GC worker could collect expired objects.

Patch #3 flush pending destroy work in module removal path, otherwise UaF
         is possible.

Patch #4 and #6 restrict the table dormant flag with basechain updates
 to fix state inconsistency in the hook registration.

Patch #5 adds missing RCU read side lock to flowtable type to avoid races
 with module removal.

* tag 'nf-24-04-04' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nf_tables: discard table flag update with pending basechain deletion
  netfilter: nf_tables: Fix potential data-race in __nft_flowtable_type_get()
  netfilter: nf_tables: reject new basechain after table flag update
  netfilter: nf_tables: flush pending destroy work before exit_net release
  netfilter: nf_tables: release mutex after nft_gc_seq_end from abort path
  netfilter: nf_tables: release batch on table validation from abort path
====================

Link: https://lore.kernel.org/r/20240404104334.1627-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
14 months agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net...
Jakub Kicinski [Thu, 4 Apr 2024 16:34:35 +0000 (09:34 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2024-04-03 (ice, idpf)

This series contains updates to ice and idpf drivers.

Dan Carpenter initializes some pointer declarations to NULL as needed for
resource cleanup on ice driver.

Petr Oros corrects assignment of VLAN operators to fix Rx VLAN filtering
in legacy mode for ice.

Joshua calls eth_type_trans() on unknown packets to prevent possible
kernel panic on idpf.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  idpf: fix kernel panic on unknown packet types
  ice: fix enabling RX VLAN filtering
  ice: Fix freeing uninitialized pointers
====================

Link: https://lore.kernel.org/r/20240403201929.1945116-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
14 months agonet/sched: act_skbmod: prevent kernel-infoleak
Eric Dumazet [Wed, 3 Apr 2024 13:09:08 +0000 (13:09 +0000)]
net/sched: act_skbmod: prevent kernel-infoleak

syzbot found that tcf_skbmod_dump() was copying four bytes
from kernel stack to user space [1].

The issue here is that 'struct tc_skbmod' has a four bytes hole.

We need to clear the structure before filling fields.

[1]
BUG: KMSAN: kernel-infoleak in instrument_copy_to_user include/linux/instrumented.h:114 [inline]
 BUG: KMSAN: kernel-infoleak in copy_to_user_iter lib/iov_iter.c:24 [inline]
 BUG: KMSAN: kernel-infoleak in iterate_ubuf include/linux/iov_iter.h:29 [inline]
 BUG: KMSAN: kernel-infoleak in iterate_and_advance2 include/linux/iov_iter.h:245 [inline]
 BUG: KMSAN: kernel-infoleak in iterate_and_advance include/linux/iov_iter.h:271 [inline]
 BUG: KMSAN: kernel-infoleak in _copy_to_iter+0x366/0x2520 lib/iov_iter.c:185
  instrument_copy_to_user include/linux/instrumented.h:114 [inline]
  copy_to_user_iter lib/iov_iter.c:24 [inline]
  iterate_ubuf include/linux/iov_iter.h:29 [inline]
  iterate_and_advance2 include/linux/iov_iter.h:245 [inline]
  iterate_and_advance include/linux/iov_iter.h:271 [inline]
  _copy_to_iter+0x366/0x2520 lib/iov_iter.c:185
  copy_to_iter include/linux/uio.h:196 [inline]
  simple_copy_to_iter net/core/datagram.c:532 [inline]
  __skb_datagram_iter+0x185/0x1000 net/core/datagram.c:420
  skb_copy_datagram_iter+0x5c/0x200 net/core/datagram.c:546
  skb_copy_datagram_msg include/linux/skbuff.h:4050 [inline]
  netlink_recvmsg+0x432/0x1610 net/netlink/af_netlink.c:1962
  sock_recvmsg_nosec net/socket.c:1046 [inline]
  sock_recvmsg+0x2c4/0x340 net/socket.c:1068
  __sys_recvfrom+0x35a/0x5f0 net/socket.c:2242
  __do_sys_recvfrom net/socket.c:2260 [inline]
  __se_sys_recvfrom net/socket.c:2256 [inline]
  __x64_sys_recvfrom+0x126/0x1d0 net/socket.c:2256
 do_syscall_64+0xd5/0x1f0
 entry_SYSCALL_64_after_hwframe+0x6d/0x75

Uninit was stored to memory at:
  pskb_expand_head+0x30f/0x19d0 net/core/skbuff.c:2253
  netlink_trim+0x2c2/0x330 net/netlink/af_netlink.c:1317
  netlink_unicast+0x9f/0x1260 net/netlink/af_netlink.c:1351
  nlmsg_unicast include/net/netlink.h:1144 [inline]
  nlmsg_notify+0x21d/0x2f0 net/netlink/af_netlink.c:2610
  rtnetlink_send+0x73/0x90 net/core/rtnetlink.c:741
  rtnetlink_maybe_send include/linux/rtnetlink.h:17 [inline]
  tcf_add_notify net/sched/act_api.c:2048 [inline]
  tcf_action_add net/sched/act_api.c:2071 [inline]
  tc_ctl_action+0x146e/0x19d0 net/sched/act_api.c:2119
  rtnetlink_rcv_msg+0x1737/0x1900 net/core/rtnetlink.c:6595
  netlink_rcv_skb+0x375/0x650 net/netlink/af_netlink.c:2559
  rtnetlink_rcv+0x34/0x40 net/core/rtnetlink.c:6613
  netlink_unicast_kernel net/netlink/af_netlink.c:1335 [inline]
  netlink_unicast+0xf4c/0x1260 net/netlink/af_netlink.c:1361
  netlink_sendmsg+0x10df/0x11f0 net/netlink/af_netlink.c:1905
  sock_sendmsg_nosec net/socket.c:730 [inline]
  __sock_sendmsg+0x30f/0x380 net/socket.c:745
  ____sys_sendmsg+0x877/0xb60 net/socket.c:2584
  ___sys_sendmsg+0x28d/0x3c0 net/socket.c:2638
  __sys_sendmsg net/socket.c:2667 [inline]
  __do_sys_sendmsg net/socket.c:2676 [inline]
  __se_sys_sendmsg net/socket.c:2674 [inline]
  __x64_sys_sendmsg+0x307/0x4a0 net/socket.c:2674
 do_syscall_64+0xd5/0x1f0
 entry_SYSCALL_64_after_hwframe+0x6d/0x75

Uninit was stored to memory at:
  __nla_put lib/nlattr.c:1041 [inline]
  nla_put+0x1c6/0x230 lib/nlattr.c:1099
  tcf_skbmod_dump+0x23f/0xc20 net/sched/act_skbmod.c:256
  tcf_action_dump_old net/sched/act_api.c:1191 [inline]
  tcf_action_dump_1+0x85e/0x970 net/sched/act_api.c:1227
  tcf_action_dump+0x1fd/0x460 net/sched/act_api.c:1251
  tca_get_fill+0x519/0x7a0 net/sched/act_api.c:1628
  tcf_add_notify_msg net/sched/act_api.c:2023 [inline]
  tcf_add_notify net/sched/act_api.c:2042 [inline]
  tcf_action_add net/sched/act_api.c:2071 [inline]
  tc_ctl_action+0x1365/0x19d0 net/sched/act_api.c:2119
  rtnetlink_rcv_msg+0x1737/0x1900 net/core/rtnetlink.c:6595
  netlink_rcv_skb+0x375/0x650 net/netlink/af_netlink.c:2559
  rtnetlink_rcv+0x34/0x40 net/core/rtnetlink.c:6613
  netlink_unicast_kernel net/netlink/af_netlink.c:1335 [inline]
  netlink_unicast+0xf4c/0x1260 net/netlink/af_netlink.c:1361
  netlink_sendmsg+0x10df/0x11f0 net/netlink/af_netlink.c:1905
  sock_sendmsg_nosec net/socket.c:730 [inline]
  __sock_sendmsg+0x30f/0x380 net/socket.c:745
  ____sys_sendmsg+0x877/0xb60 net/socket.c:2584
  ___sys_sendmsg+0x28d/0x3c0 net/socket.c:2638
  __sys_sendmsg net/socket.c:2667 [inline]
  __do_sys_sendmsg net/socket.c:2676 [inline]
  __se_sys_sendmsg net/socket.c:2674 [inline]
  __x64_sys_sendmsg+0x307/0x4a0 net/socket.c:2674
 do_syscall_64+0xd5/0x1f0
 entry_SYSCALL_64_after_hwframe+0x6d/0x75

Local variable opt created at:
  tcf_skbmod_dump+0x9d/0xc20 net/sched/act_skbmod.c:244
  tcf_action_dump_old net/sched/act_api.c:1191 [inline]
  tcf_action_dump_1+0x85e/0x970 net/sched/act_api.c:1227

Bytes 188-191 of 248 are uninitialized
Memory access of size 248 starts at ffff888117697680
Data copied to user address 00007ffe56d855f0

Fixes: 86da71b57383 ("net_sched: Introduce skbmod action")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/20240403130908.93421-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
14 months agonet: usb: ax88179_178a: avoid the interface always configured as random address
Jose Ignacio Tornos Martinez [Wed, 3 Apr 2024 13:21:58 +0000 (15:21 +0200)]
net: usb: ax88179_178a: avoid the interface always configured as random address

After the commit d2689b6a86b9 ("net: usb: ax88179_178a: avoid two
consecutive device resets"), reset is not executed from bind operation and
mac address is not read from the device registers or the devicetree at that
moment. Since the check to configure if the assigned mac address is random
or not for the interface, happens after the bind operation from
usbnet_probe, the interface keeps configured as random address, although the
address is correctly read and set during open operation (the only reset
now).

In order to keep only one reset for the device and to avoid the interface
always configured as random address, after reset, configure correctly the
suitable field from the driver, if the mac address is read successfully from
the device registers or the devicetree. Take into account if a locally
administered address (random) was previously stored.

cc: stable@vger.kernel.org # 6.6+
Fixes: d2689b6a86b9 ("net: usb: ax88179_178a: avoid two consecutive device resets")
Reported-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Jose Ignacio Tornos Martinez <jtornosm@redhat.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240403132158.344838-1-jtornosm@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
14 months agonvme-fc: rename free_ctrl callback to match name pattern
Daniel Wagner [Thu, 4 Apr 2024 14:41:31 +0000 (16:41 +0200)]
nvme-fc: rename free_ctrl callback to match name pattern

Rename nvme_fc_nvme_ctrl_freed to nvme_fc_free_ctrl to match the name
pattern for the callback.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Daniel Wagner <dwagner@suse.de>
Signed-off-by: Keith Busch <kbusch@kernel.org>
14 months agonvmet-fc: move RCU read lock to nvmet_fc_assoc_exists
Daniel Wagner [Thu, 4 Apr 2024 14:41:30 +0000 (16:41 +0200)]
nvmet-fc: move RCU read lock to nvmet_fc_assoc_exists

The RCU lock is only needed for the lookup loop and not for
list_ad_tail_rcu call. Thus move it down the call chain into
nvmet_fc_assoc_exists.

While at it also fix the name typo of the function.

Signed-off-by: Daniel Wagner <dwagner@suse.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Keith Busch <kbusch@kernel.org>
14 months agonvmet: implement unique discovery NQN
Hannes Reinecke [Wed, 3 Apr 2024 11:31:14 +0000 (13:31 +0200)]
nvmet: implement unique discovery NQN

Unique discovery NQNs allow to differentiate between discovery
services from (typically physically separate) NVMe-oF subsystems.
This is required for establishing secured connections as otherwise
the credentials won't be unique and the integrity of the connection
cannot be guaranteed.
This patch adds a configfs attribute 'discovery_nqn' in the 'nvmet'
configfs directory to specify the unique discovery NQN.

Signed-off-by: Hannes Reinecke <hare@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
14 months agonvme: don't create a multipath node for zero capacity devices
Christoph Hellwig [Wed, 3 Apr 2024 12:47:17 +0000 (14:47 +0200)]
nvme: don't create a multipath node for zero capacity devices

Apparently there are nvme controllers around that report namespaces
in the namespace list which have zero capacity.  Return -ENXIO instead
of -ENODEV from nvme_update_ns_info_block so we don't create a hidden
multipath node for these namespaces but entirely ignore them.

Fixes: 46e7422cda84 ("nvme: move common logic into nvme_update_ns_info")
Reported-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Tested-by: Nilay Shroff <nilay@linux.ibm.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
14 months agox86/mce: Make sure to grab mce_sysfs_mutex in set_bank()
Borislav Petkov (AMD) [Wed, 13 Mar 2024 13:48:27 +0000 (14:48 +0100)]
x86/mce: Make sure to grab mce_sysfs_mutex in set_bank()

Modifying a MCA bank's MCA_CTL bits which control which error types to
be reported is done over

  /sys/devices/system/machinecheck/
  ├── machinecheck0
  │   ├── bank0
  │   ├── bank1
  │   ├── bank10
  │   ├── bank11
  ...

sysfs nodes by writing the new bit mask of events to enable.

When the write is accepted, the kernel deletes all current timers and
reinits all banks.

Doing that in parallel can lead to initializing a timer which is already
armed and in the timer wheel, i.e., in use already:

  ODEBUG: init active (active state 0) object: ffff888063a28000 object
  type: timer_list hint: mce_timer_fn+0x0/0x240 arch/x86/kernel/cpu/mce/core.c:2642
  WARNING: CPU: 0 PID: 8120 at lib/debugobjects.c:514
  debug_print_object+0x1a0/0x2a0 lib/debugobjects.c:514

Fix that by grabbing the sysfs mutex as the rest of the MCA sysfs code
does.

Reported by: Yue Sun <samsun1006219@gmail.com>
Reported by: xingwei lee <xrivendell7@gmail.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Cc: <stable@kernel.org>
Link: https://lore.kernel.org/r/CAEkJfYNiENwQY8yV1LYJ9LjJs%2Bx_-PqMv98gKig55=2vbzffRw@mail.gmail.com
14 months agogpio: cdev: check for NULL labels when sanitizing them for irqs
Bartosz Golaszewski [Thu, 4 Apr 2024 09:33:27 +0000 (11:33 +0200)]
gpio: cdev: check for NULL labels when sanitizing them for irqs

We need to take into account that a line's consumer label may be NULL
and not try to kstrdup() it in that case but rather pass the NULL
pointer up the stack to the interrupt request function.

To that end: let make_irq_label() return NULL as a valid return value
and use ERR_PTR() instead to signal an allocation failure to callers.

Cc: stable@vger.kernel.org
Fixes: b34490879baa ("gpio: cdev: sanitize the label before requesting the interrupt")
Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
Closes: https://lore.kernel.org/lkml/20240402093534.212283-1-naresh.kamboju@linaro.org/
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Tested-by: Anders Roxell <anders.roxell@linaro.org>
14 months agoSUNRPC: Fix a slow server-side memory leak with RPC-over-TCP
Chuck Lever [Wed, 3 Apr 2024 14:36:25 +0000 (10:36 -0400)]
SUNRPC: Fix a slow server-side memory leak with RPC-over-TCP

Jan Schunk reports that his small NFS servers suffer from memory
exhaustion after just a few days. A bisect shows that commit
e18e157bb5c8 ("SUNRPC: Send RPC message on TCP with a single
sock_sendmsg() call") is the first bad commit.

That commit assumed that sock_sendmsg() releases all the pages in
the underlying bio_vec array, but the reality is that it doesn't.
svc_xprt_release() releases the rqst's response pages, but the
record marker page fragment isn't one of those, so it is never
released.

This is a narrow fix that can be applied to stable kernels. A
more extensive fix is in the works.

Reported-by: Jan Schunk <scpcom@gmx.de>
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=218671
Fixes: e18e157bb5c8 ("SUNRPC: Send RPC message on TCP with a single sock_sendmsg() call")
Cc: Alexander Duyck <alexander.duyck@gmail.com>
Cc: Jakub Kacinski <kuba@kernel.org>
Cc: David Howells <dhowells@redhat.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
14 months agodrm/xe: Use ordered wq for preempt fence waiting
Matthew Brost [Mon, 1 Apr 2024 22:19:11 +0000 (15:19 -0700)]
drm/xe: Use ordered wq for preempt fence waiting

Preempt fences can sleep waiting for an exec queue suspend operation to
complete. If the system_unbound_wq is used for waiting and the number of
waiters exceeds max_active this will result in other users of the
system_unbound_wq getting starved. Use a device private work queue for
preempt fences to avoid starvation of the system_unbound_wq.

Even though suspend operations can complete out-of-order, all suspend
operations within a VM need to complete before the preempt rebind worker
can start. With that, use a device private ordered wq for preempt fence
waiting.

v2:
 - Add comment about cleanup on failure (Matt R)
 - Update commit message (Lucas)

Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240401221913.139672-2-matthew.brost@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
(cherry picked from commit 37c15c4aae1fe3f67efd2641db8d8c25c2d524ab)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
14 months agodrm/xe: Move vma rebinding to the drm_exec locking loop
Thomas Hellström [Wed, 27 Mar 2024 09:11:36 +0000 (10:11 +0100)]
drm/xe: Move vma rebinding to the drm_exec locking loop

Rebinding might allocate page-table bos, causing evictions.
To support blocking locking during these evictions,
perform the rebinding in the drm_exec locking loop.

Also Reserve fence slots where actually needed rather than trying to
predict how many fence slots will be needed over a complete
wound-wait transaction.

v2:
- Remove a leftover call to xe_vm_rebind() (Matt Brost)
- Add a helper function xe_vm_validate_rebind() (Matt Brost)
v3:
- Add comments and squash with previous patch (Matt Brost)

Fixes: 24f947d58fe5 ("drm/xe: Use DRM GPUVM helpers for external- and evicted objects")
Fixes: 29f424eb8702 ("drm/xe/exec: move fence reservation")
Cc: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240327091136.3271-5-thomas.hellstrom@linux.intel.com
(cherry picked from commit 7ee7dd6f301341d5b1204fc19fa620d7f7f7e90d)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
14 months agodrm/xe: Make TLB invalidation fences unordered
Thomas Hellström [Wed, 27 Mar 2024 09:11:35 +0000 (10:11 +0100)]
drm/xe: Make TLB invalidation fences unordered

They can actually complete out-of-order, so allocate a unique
fence context for each fence.

Fixes: 5387e865d90e ("drm/xe: Add TLB invalidation fence after rebinds issued from execs")
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <stable@vger.kernel.org> # v6.8+
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240327091136.3271-4-thomas.hellstrom@linux.intel.com
(cherry picked from commit 0453f1757501df2e82b66b3183a24bba5a6f8fa3)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
14 months agodrm/xe: Rework rebinding
Thomas Hellström [Wed, 27 Mar 2024 09:11:34 +0000 (10:11 +0100)]
drm/xe: Rework rebinding

Instead of handling the vm's rebind fence separately,
which is error prone if they are not strictly ordered,
attach rebind fences as kernel fences to the vm's resv.

Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <stable@vger.kernel.org> # v6.8+
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240327091136.3271-3-thomas.hellstrom@linux.intel.com
(cherry picked from commit 5a091aff50b780ae29c7faf70a7a6c21c98a54c4)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
14 months agodrm/xe: Use ring ops TLB invalidation for rebinds
Thomas Hellström [Wed, 27 Mar 2024 09:11:33 +0000 (10:11 +0100)]
drm/xe: Use ring ops TLB invalidation for rebinds

For each rebind we insert a GuC TLB invalidation and add a
corresponding unordered TLB invalidation fence. This might
add a huge number of TLB invalidation fences to wait for so
rather than doing that, defer the TLB invalidation to the
next ring ops for each affected exec queue. Since the TLB
is invalidated on exec_queue switch, we need to invalidate
once for each affected exec_queue.

v2:
- Simplify if-statements around the tlb_flush_seqno.
  (Matthew Brost)
- Add some comments and asserts.

Fixes: 5387e865d90e ("drm/xe: Add TLB invalidation fence after rebinds issued from execs")
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <stable@vger.kernel.org> # v6.8+
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240327091136.3271-2-thomas.hellstrom@linux.intel.com
(cherry picked from commit 4fc4899e86f7afbd09f4bcb899f0fc57e0296e62)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
14 months agoASoC: SOF: amd: fix for false dsp interrupts
Vijendar Mukunda [Thu, 4 Apr 2024 04:17:13 +0000 (09:47 +0530)]
ASoC: SOF: amd: fix for false dsp interrupts

Before ACP firmware loading, DSP interrupts are not expected.
Sometimes after reboot, it's observed that before ACP firmware is loaded
false DSP interrupt is reported.
Registering the interrupt handler before acp initialization causing false
interrupts sometimes on reboot as ACP reset is not applied.
Correct the sequence by invoking acp initialization sequence prior to
registering interrupt handler.

Fixes: 738a2b5e2cc9 ("ASoC: SOF: amd: Add IPC support for ACP IP block")
Signed-off-by: Vijendar Mukunda <Vijendar.Mukunda@amd.com>
Link: https://msgid.link/r/20240404041717.430545-1-Vijendar.Mukunda@amd.com
Signed-off-by: Mark Brown <broonie@kernel.org>
14 months agonet: dsa: sja1105: Fix parameters order in sja1110_pcs_mdio_write_c45()
Christophe JAILLET [Tue, 2 Apr 2024 18:33:56 +0000 (20:33 +0200)]
net: dsa: sja1105: Fix parameters order in sja1110_pcs_mdio_write_c45()

The definition and declaration of sja1110_pcs_mdio_write_c45() don't have
parameters in the same order.

Knowing that sja1110_pcs_mdio_write_c45() is used as a function pointer
in 'sja1105_info' structure with .pcs_mdio_write_c45, and that we have:

   int (*pcs_mdio_write_c45)(struct mii_bus *bus, int phy, int mmd,
  int reg, u16 val);

it is likely that the definition is the one to change.

Found with cppcheck, funcArgOrderDifferent.

Fixes: ae271547bba6 ("net: dsa: sja1105: C45 only transactions for PCS")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Michael Walle <mwalle@kernel.org>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/ff2a5af67361988b3581831f7bd1eddebfb4c48f.1712082763.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Paolo Abeni <pabeni@redhat.com>