spapr: handle cpu core unplug via hotplug handler chain

Factor out cpu core unplug into separate function from
spapr_core_release(). Then use generic hotplug_handler_unplug() to trigger
cpu core unplug, which would call spapr_machine_device_unplug() ->
spapr_core_unplug() in the end.

This way unplug operation is not buried in spapr internals and located
in the same place like in other targets, following similar
logic/call chain across targets.

Acked-by: Igor Mammedov <imammedo@redhat.com>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: handle pc-dimm unplug via hotplug handler chain

Factor out memory unplug into separate function from spapr_lmb_release().
Then use generic hotplug_handler_unplug() to trigger memory unplug,
which will call spapr_machine_device_unplug() -> spapr_memory_unplug()
in the end.

This way unplug operation is not buried in lmb internals and located in
the same place like in other targets, following similar logic/call chain
across targets.

Acked-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: introduce machine unplug handler

We'll be handling unplug of e.g. CPUs and PCDIMMs  via the general
hotplug handler soon, so let's add that handler function.

Acked-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: move memory hotplug support check into spapr_memory_pre_plug()

Let's finish cleaning up the hotplug handler. This check can be
performed in the pre_plug code as the very first thing.

Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: move lookup of the node into spapr_memory_plug()

Let's clean the hotplug handler up by moving lookup of the node into
the function where it is actually being used.

Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: no need to verify the node

The node property can always be queried and the value has already been
verified in pc_dimm_realize().

Acked-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Allow PIR read in privileged mode

According to PowerISA, the PIR register should be readable in privileged
mode also, not only in hypervisor privileged mode.

PowerISA 3.0 - 4.3.3 Processor Identification Register

"Read access to the PIR is privileged; write access is not provided."

Figure 18 in section 4.4.4 explicitly confirms that mfspr PIR is privileged
and doesn't require hypervisor state.

Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Alexander Graf <agraf@suse.de>
Cc: qemu-ppc@nongnu.org
Signed-off-by: Leandro Lupori <leandro.lupori@gmail.com>
Reviewed-by: Jose Ricardo Ziviani <joserz@linux.ibm.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

ppc4xx_i2c: Clean up and improve error logging

Make it more readable by converting register indexes to decimal
(avoids lot of superfluous 0x0) and distinguish errors caused by
accessing non-existent vs. unimplemented registers.
No functional change.

Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: extend eieio for POWER9

POWER9 introduced a new variant of the eieio instruction using bit 6
as a hint to tell the CPU it is a store-forwarding barrier.

The usage of this eieio extension was recently added in Linux 4.17
which activated the "support for a store forwarding barrier at kernel
entry/exit".

Unfortunately, it is not possible to insert this new eieio instruction
without considerable change in ppc_tr_translate_insn(). So instead we
loosen the QEMU eieio instruction mask and modify the gen_eieio()
helper to test for bit6. On non-POWER9 CPUs, the bit6 is just ignored
but a warning is emitted as this is not an instruction software should
be using.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

mos6522: convert VMSTATE_TIMER_PTR_TEST to VMSTATE_TIMER_PTR

The timers are configured in the mos6522 init function and therefore will
always exist, so the function can never return false.

Peter also pointed out that this is the only remaining user of
VMSTATE_TIMER_PTR_TEST in the codebase, so we might as well just convert it
over to VMSTATE_TIMER_PTR and remove mos6522_timer_exist() as it is no
longer required.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

mos6522: move timer frequency initialisation to mos6522_reset

The 6522 VIA timer frequency cannot be set by altering registers within the
device itself and hence it is a fixed property of the machine.

Move the initialisation of the timer frequency to the mos6522 reset function
and ensure that any subclasses always call the parent reset function so that
it isn't required to store the timer frequency within vmstate_mos6522_timer
itself.

By moving the frequency initialisation to the device reset function then we
find that the realize function for both mos6522 and mos6522_cuda becomes
obsolete and can simply be removed.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

cuda: embed mos6522_cuda device directly rather than using QOM object link

Examining the migration stream it can be seen that the mos6522 device state is
being stored separately rather than as part of the CUDA device which is
incorrect (and likely to cause issues if another mos6522 device is added to
the machine).

Resolve this by embedding the mos6522_cuda device directly within the CUDA
device rather than using a QOM object link to reference the device separately.

Note that we also bump the version in vmstate_cuda to reflect this change: this
isn't particularly important for the moment as the Mac machine migration isn't
100% reliable due to issues migrating the timebase under TCG.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

mos6522: fix vmstate_mos6522_timer version in vmstate_mos6522

This was accidentally introduced when extracting the 6522 VIA functionality
from the CUDA device, and prevents loadvm from completing successfully.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

ppc: add missing FW_CFG_PPC_NVRAM_FLAT definition

This is used in OpenBIOS to define the memory layout of the NVRAM device. Whilst
currently left at its default value, add the missing definition to ensure it is
reserved.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

ppc: remove obsolete macio_init() definition from mac.h

Commits b6712ea391 removed the macio_init() function but missed the header
prototype in mac.h. Remove it since it is no longer needed.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

ppc: remove obsolete pci_pmac_init() definitions from mac.h

Commits 7b19318bee and 8ce3f743c7 removed the pci_pmac_init() and
pci_pmac_u3_init() functions but missed the header prototypes in mac.h. Remove
them since they are no longer needed.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

hw/misc/mos6522: Add trailing '\n' to qemu_log() calls

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

MAINTAINERS: Add entries for the MOS6522 VIA device

Introduced in 51f233ec92c.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

MAINTAINERS: Add an entry for the MacIO device headers

Missed while moved in 7092e84d42b.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

osdep: powerpc64 align memory to allow 2MB radix THP page tables

This allows KVM with the Book3S radix MMU mode to take advantage of
THP and install larger pages in the partition scope page tables (the
host translation).

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr/vio: deprecate the "irq" property

VIO devices have an "irq" property that can be used by the sPAPR IRQ
allocator as an IRQ number hint. But it is not set in QEMU nor in
libvirt. It brings unnecessary complexity to the underlying layers
managing the IRQ number space and it is in full opposition with the
new static IRQ allocator we want to introduce in sPAPR.

Let's deprecate it to simplify the spapr_irq_alloc routine in the
future.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
[dwg: Check qtest_enabled() to suppress bogus warnings from make check]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Allow privileged access to SPR_PCR

The powerpc Linux kernel[1] and skiboot firmware[2] recently gained changes
that cause the Processor Compatibility Register (PCR) SPR to be cleared.

These changes cause Linux to fail to boot on the Qemu powernv machine
with an error:

 Trying to write privileged spr 338 (0x152) at 0000000030017f0c

With this patch Qemu makes this register available as a hypervisor
privileged register.

Note that bits set in this register disable features of the processor.
Currently the only register state that is supported is when the register
is zeroed (enable all features). This is sufficient for guests to
once again boot.

[1] https://lkml.kernel.org/r/20180518013742.24095-1-mikey@neuling.org
[2] https://patchwork.ozlabs.org/patch/915932/

Signed-off-by: Joel Stanley <joel@jms.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Factor out the parsing in kvmppc_get_cpu_characteristics()

Factor out the parsing of struct kvm_ppc_cpu_char in
kvmppc_get_cpu_characteristics() into a separate function for each cap
for simplicity.

Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

prep: fix keyboard for the 40p machine

Commit 72d3d8f052 "hw/isa/superio: Add a keyboard/mouse controller (8042)"
added an 8042 keyboard device to the PC87312 superio device to replace that
being used by the prep machine.

Unfortunately this commit didn't do the same for the 40p machine which broke
the keyboard by registering two 8042 keyboard devices at the same address.

Resolve this by similarly removing the 8042 keyboard from the 40p machine as
done for the prep machine in commit 72d3d8f052.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Hervé Poussineau <hpoussin@reactos.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

40p: remove pci_allow_0_address = true from 40p machine class

The Linux sandalfoot zImage has an initialisation process which resets the
VGA controller by setting all the BAR addresses to zero to access the VGA
ioports at their legacy addresses.

Unfortunately setting the framebuffer BAR to address 0 makes the framebuffer
memory overlap the internal VGA memory causing accesses to fail, and so
prevents the kernel from switching successfully to text mode.

Since OpenHackWare configures the framebuffer BAR address outside of the legacy
VGA internal memory space, remove pci_allow_0_address from the 40p machine class
which causes the BAR reprogramming to zero to fail and so the VGA internal
memory can be accessed correctly again.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Use proper logging function for possible guest errors

fprintf() and qemu_log_separate() are frowned upon these days for printing
logging information in QEMU. Accessing the wrong SPRs indicates wrong guest
behaviour in most cases, and we've got a proper way to log such situations,
which is the qemu_log_mask(LOG_GUEST_ERROR, ...) function. So use this
function now for logging the bad SPR accesses instead.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

hw/ppc/spapr_drc: Replace error_setg(&error_abort) by error_report() + abort()

Use error_report() + abort() instead of error_setg(&error_abort),
as suggested by the "qapi/error.h" documentation:

    Please don't error_setg(&error_fatal, ...), use error_report() and
    exit(), because that's more obvious.
    Likewise, don't error_setg(&error_abort, ...), use assert().

Use abort() instead of the suggested assert() because the error message
already got displayed.

Suggested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

uninorth: remove token register from uninorth device

>From observation of various OS sources it can be seen that the token register
introduced in 4e46dcdbd3 "PPC: Newworld: Add uninorth token register" is not
required, since the only register currently implemented is the uninorth hardware
version which is read-only.

Remove the token register implementation and instead return the uninorth
version corresponding to the hardware.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

macio: add trace-events to timer device

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

ppc440_pcix: Fix a typo in setting a register (Coverity CID1390577)

Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2018-06-11' into staging

Block patches:
- Various bug fixes
- Removal of qemu-img convert's deprecated -s option
- qemu-io now exits with an error when a command failed

# gpg: Signature made Mon 11 Jun 2018 15:23:42 BST
# gpg:                using RSA key F407DB0061D5CF40
# gpg: Good signature from "Max Reitz <mreitz@redhat.com>"
# Primary key fingerprint: 91BE B60A 30DB 3E88 57D1  1829 F407 DB00 61D5 CF40

* remotes/maxreitz/tags/pull-block-2018-06-11: (29 commits)
  iotests: Add case for a corrupted inactive image
  qcow2: Do not mark inactive images corrupt
  block: Make bdrv_is_writable() public
  throttle: Fix crash on reopen
  block/qcow2-bitmap: fix free_bitmap_clusters
  qemu-img: Remove deprecated -s snapshot_id_or_name option
  iotests: Fix 219's timing
  iotests: improve pause_job
  iotests: Test post-backing convert target behavior
  qemu-img: Special post-backing convert handling
  iotests: Add test for rebasing with relative paths
  qemu-img: Resolve relative backing paths in rebase
  iotests: Let 216 make use of qemu-io's exit code
  iotests.py: Add qemu_io_silent
  qemu-io: Exit with error when a command failed
  qemu-io: Let command functions return error code
  qemu-io: Drop command functions' return values
  iotests: Repairing error during snapshot deletion
  qcow2: Repair OFLAG_COPIED when fixing leaks
  iotests: Rework 113
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

iotests: Add case for a corrupted inactive image

Reviewed-by: John Snow <jsnow@redhat.com>
Tested-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180606193702.7113-4-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Do not mark inactive images corrupt

When signaling a corruption on a read-only image, qcow2 already makes
fatal events non-fatal (i.e., they will not result in the image being
closed, and the image header's corrupt flag will not be set).  This is
necessary because we cannot set the corrupt flag on read-only images,
and it is possible because further corruption of read-only images is
impossible.

Inactive images are effectively read-only, too, so we should do the same
for them.  bdrv_is_writable() can tell us whether an image can actually
be written to, so use its result instead of !bs->read_only.

(Otherwise, the assert(!(bs->open_flags & BDRV_O_INACTIVE)) in
bdrv_co_pwritev() will fail, crashing qemu.)

Cc: qemu-stable@nongnu.org
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180606193702.7113-3-mreitz@redhat.com
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

block: Make bdrv_is_writable() public

This is a useful function for the whole block layer, so make it public.
At the same time, users outside of block.c probably do not need to make
use of the reopen functionality, so rename the current function to
bdrv_is_writable_after_reopen() create a new bdrv_is_writable() function
that just passes NULL to it for the reopen queue.

Cc: qemu-stable@nongnu.org
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180606193702.7113-2-mreitz@redhat.com
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

throttle: Fix crash on reopen

The throttle block filter can be reopened, and with this it is
possible to change the throttle group that the filter belongs to.

The way the code does that is the following:

  - On throttle_reopen_prepare(): create a new ThrottleGroupMember
    and attach it to the new throttle group.

  - On throttle_reopen_commit(): detach the old ThrottleGroupMember,
    delete it and replace it with the new one.

The problem with this is that by replacing the ThrottleGroupMember the
previous value of io_limits_disabled is lost, causing an assertion
failure in throttle_co_drain_end().

This problem can be reproduced by reopening a throttle node:

   $QEMU -monitor stdio
   -object throttle-group,id=tg0,x-iops-total=1000 \
   -blockdev node-name=hd0,driver=qcow2,file.driver=file,file.filename=hd.qcow2 \
   -blockdev node-name=root,driver=throttle,throttle-group=tg0,file=hd0,read-only=on

   (qemu) block_stream root
   block/throttle.c:214: throttle_co_drain_end: Assertion `tgm->io_limits_disabled' failed.

Since we only want to change the throttle group on reopen there's no
need to create a ThrottleGroupMember and discard the old one. It's
easier if we simply detach it from its current group and attach it to
the new one.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 20180608151536.7378-1-berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

block/qcow2-bitmap: fix free_bitmap_clusters

This assert may fail, because bitmap_table is not initialized. Just
drop it, as it's obvious, that bitmap_table_load sets bitmap_table
parameter only when returning zero.

Reported-by: Pavel Butsykin <pbutsykin@virtuozzo.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-id: 20180608101225.2575-1-vsementsov@virtuozzo.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-img: Remove deprecated -s snapshot_id_or_name option

It has been marked as deprecated since QEMU v2.0 already, so it
is time now to finally remove it.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-id: 1528288551-31641-1-git-send-email-thuth@redhat.com
Reviewed-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Fix 219's timing

219 has two issues that may lead to sporadic failure, both of which are
the result of issuing query-jobs too early after a job has been
modified.  This can then lead to different results based on whether the
modification has taken effect already or not.

First, query-jobs is issued right after the job has been created.
Besides its current progress possibly being in any random state (which
has already been taken care of), its total progress too is basically
arbitrary, because the job may not yet have been able to determine it.
This patch addresses this by just filtering the total progress, like
what has been done for the current progress already.  However, for more
clarity, the filtering is changed to replace the values by a string
'FILTERED' instead of deleting them.

Secondly, query-jobs is issued right after a job has been resumed.  The
job may or may not yet have had the time to actually perform any I/O,
and thus its current progress may or may not have advanced.  To make
sure it has indeed advanced (which is what the reference output already
assumes), keep querying it until it has.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180606190628.8170-1-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: improve pause_job

It's possible, that job was finished during waiting. In this case we
will see error message "Timeout waiting for job to pause" which is not
very informative. So, let's check during waiting iteration that the job
exists.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-id: 20180601115923.17159-1-vsementsov@virtuozzo.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Test post-backing convert target behavior

This adds a test case to 122 for what happens when you convert to a
target with a backing file that is shorter than the target, and the
image format does not support efficient zero writes (as is the case with
qcow2 v2).

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180501165750.19242-3-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-img: Special post-backing convert handling

Currently, qemu-img convert writes zeroes when it reads zeroes.
Sometimes it does not because the target is initialized to zeroes
anyway, so we do not need to overwrite (and thus potentially allocate)
it.  This is never the case for targets with backing files, though.  But
even they may have an area that is initialized to zeroes, and that is
the area past the end of the backing file (if that is shorter than the
overlay).

So if the target format's unallocated blocks are zero and there is a gap
between the target's backing file's end and the target's end, we do not
have to explicitly write zeroes there.

Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1527898
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180501165750.19242-2-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Add test for rebasing with relative paths

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509182002.8044-3-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-img: Resolve relative backing paths in rebase

Currently, rebase interprets a relative path for the new backing image
as follows:
(1) Open the new backing image with the given relative path (thus relative to
    qemu-img's working directory).
(2) Write it directly into the overlay's backing path field (thus
    relative to the overlay).

If the overlay is not in qemu-img's working directory, both will be
different interpretations, which may either lead to an error somewhere
(either rebase fails because it cannot open the new backing image, or
your overlay becomes unusable because its backing path does not point to
a file), or, even worse, it may result in your rebase being performed
for a different backing file than what your overlay will point to after
the rebase.

Fix this by interpreting the target backing path as relative to the
overlay, like qemu-img does everywhere else.

Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1569835
Cc: qemu-stable@nongnu.org
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180509182002.8044-2-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Let 216 make use of qemu-io's exit code

As a showcase of how you can use qemu-io's exit code to determine
success or failure (same for qemu-img), this test is changed to use
qemu_io_silent() instead of qemu_io(), and to assert the exit code
instead of logging the filtered result.

One real advantage of this is that in case of an error, you get a
backtrace that helps you locate the issue in the test file quickly.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509194302.21585-6-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests.py: Add qemu_io_silent

With qemu-io now returning a useful exit code, some tests may find it
sufficient to just query that instead of logging (and filtering) the
whole output.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509194302.21585-5-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-io: Exit with error when a command failed

Currently, qemu-io basically always returns success when it gets to
interactive mode (so once the whole command line has been parsed; even
before the commands on the command line are interpreted).  That is not
very useful.

This patch makes qemu-io return failure when any of the executed
commands failed.

Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1519617
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509194302.21585-4-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-io: Let command functions return error code

This is basically what everything else in the qemu code base does, so we
can do it here, too.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509194302.21585-3-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-io: Drop command functions' return values

For qemu-io, a function returns an integer with two possible values: 0
for "qemu-io may continue execution", or 1 for "qemu-io should exit".
However, there is only a single command that returns 1, and that is
"quit".

So let's turn this case into a global variable instead so we can make
better use of the return value in a later patch.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509194302.21585-2-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Repairing error during snapshot deletion

This adds a test for an I/O error during snapshot deletion, and maybe
more importantly, for how to repair the resulting image.  If the
snapshot has been deleted before the error occurs, the only negative
result will be leaked clusters -- and those should be repairable with
qemu-img check -r leaks.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509200059.31125-3-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Repair OFLAG_COPIED when fixing leaks

Repairing OFLAG_COPIED is usually safe because it is done after the
refcounts have been repaired.  Therefore, it we did not find anyone else
referencing a data or L2 cluster, it makes no sense to not set
OFLAG_COPIED -- and the other direction (clearing OFLAG_COPIED) is
always safe, anyway, it may just induce leaks.

Furthermore, if OFLAG_COPIED is actually consistent with a wrong (leaky)
refcount, we will decrement the refcount with -r leaks, but OFLAG_COPIED
will then be wrong.  qemu-img check should not produce images that are
more corrupted afterwards then they were before.

Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1527085
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509200059.31125-2-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Rework 113

This test case has been broken since 398e6ad014df261d (roughly half a
year).  qemu-img amend requires its output image to be R/W, so it opens
it as such; the node is then turned into an read-only node automatically
which is now accompanied by a warning, however.  This warning has not
been part of the reference output.

For one thing, this warning shows that we cannot keep the test case as
it is.  We would need a format that has no create_opts but that does
have write support -- we do not have such a format, though.

Another thing is that qemu now actually checks whether an image format
supports amendment instead of whether it has create_opts (since the
former always implies the latter).  So we can now use any format that
does not support amendment (even if it supports creation) and thus test
the same code path.

The reason nobody has noticed the breakage until now of course is the
fact that nobody runs the iotests for nbd+bochs.  There actually was
never any reason to set the protocol to "nbd" but because that was
technically correct; functionally it made no difference.  So that is the
first thing we are going to change: Make the protocol "file" instead so
that people might actually notice breakage here.

Secondly, now that bochs no longer works for the amend test case, we
have to change the format there anyway.  Set let us just bend the truth
a bit, declare this test a raw test.  In fact, that does not even
concern the bochs test cases, other than the output now reading 'bochs'
instead of 'IMGFMT'.

So with this test now being a raw test, we can rework the amend test
case to use raw instead.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20180509210023.20283-8-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Test help option for unsupporting formats

This adds test cases to 082 for qemu-img create/convert/amend "-o help"
on formats that do not support creation or amendment, respectively.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509210023.20283-7-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-img: Recognize no creation support in -o help

The only users of print_block_option_help() are qemu-img create and
qemu-img convert for the output image, so this function is always used
for image creation (it used to be used for amendment also, but that is
no longer the case).

So if image creation is not supported by either the format or the
protocol, there is no need to print any option description, because the
user cannot create an image like this anyway.

This also fixes an assertion failure:

    $ qemu-img create -f bochs -o help
    Supported options:
    qemu-img: util/qemu-option.c:219:
    qemu_opts_print_help: Assertion `list' failed.
    [1]    24831 abort (core dumped)  qemu-img create -f bochs -o help

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509210023.20283-6-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-img: Add print_amend_option_help()

The more generic print_block_option_help() function is not really
suitable for qemu-img amend, for a couple of reasons:
(1) We do not need to append the protocol-level options, as amendment
    happens only on one node and does not descend downwards to its
    children.
(2) print_block_option_help() says those options are "supported".  For
    option amendment, we do not really know that.  So this new function
    explicitly says that those options are the creation options, and not
    all of them may be supported.
(3) If the driver does not support option amendment, we should not print
    anything (except for an error message that amendment is not
    supported).

Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1537956
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509210023.20283-5-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-option: Pull out "Supported options" print

It really is up to the caller to decide what this list of options means.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509210023.20283-4-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

block: Add Error parameter to bdrv_amend_options

Looking at the qcow2 code that is riddled with error_report() calls,
this is really how it should have been from the start.

Along the way, turn the target_version/current_version comparisons at
the beginning of qcow2_downgrade() into assertions (the caller has to
make sure these conditions are met), and rephrase the error message on
using compat=1.1 to get refcount widths other than 16 bits.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180509210023.20283-3-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

qemu-img: Amendment support implies create_opts

Instead of checking whether a driver has a non-NULL create_opts we
should check whether it supports image amendment in the first place.  If
it does, it must have create_opts.

On the other hand, if it does not have create_opts (so it does not
support amendment either), the error message "does not support any
options" is a bit useless.  Stating clearly that the driver has no
amendment support whatsoever is probably better.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180509210023.20283-2-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Add creation test to 153

This patch adds a test case to 153 which tries to overwrite an image
(using qemu-img create) while it is in use.  Without the original user
explicitly sharing the necessary permissions (writing and truncation),
this should not be allowed.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Message-id: 20180509215336.31304-4-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

block/file-posix: File locking during creation

When creating a file, we should take the WRITE and RESIZE permissions.
We do not need either for the creation itself, but we do need them for
clearing and resizing it.  So we can take the proper permissions by
replacing O_TRUNC with an explicit truncation to 0, and by taking the
appropriate file locks between those two steps.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180509215336.31304-3-mreitz@redhat.com
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

block/file-posix: Pass FD to locking helpers

raw_apply_lock_bytes() and raw_check_lock_bytes() currently take a
BDRVRawState *, but they only use the lock_fd field.  During image
creation, we do not have a BDRVRawState, but we do have an FD; so if we
want to reuse the functions there, we should modify them to receive only
the FD.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Message-id: 20180509215336.31304-2-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

Merge remote-tracking branch 'remotes/vivier/tags/m68k-for-3.0-pull-request' into staging

Convert to TranslatorOps

I've updated the series to fix conflicts with:

21528149eb target/m68k: Add trailing '\n' to qemu_log() call
07ea28b418 tcg: Pass tb and index to tcg_gen_exit_tb separately

# gpg: Signature made Mon 11 Jun 2018 11:48:52 BST
# gpg:                using RSA key F30C38BD3F2FBE3C
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>"
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>"
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>"
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier/tags/m68k-for-3.0-pull-request:
  target/m68k: Merge disas_m68k_insn into m68k_tr_translate_insn
  target/m68k: Improve ending TB at page boundaries
  target/m68k: Convert to TranslatorOps
  target/m68k: Convert to DisasContextBase
  target/m68k: Rename DISAS_UPDATE and gen_lookup_tb
  target/m68k: Use lookup_and_goto_tb for DISAS_JUMP
  target/m68k: Remove DISAS_JUMP_NEXT as unused
  target/m68k: Replace DISAS_TB_JUMP with DISAS_NORETURN
  target/m68k: Use DISAS_NORETURN for exceptions

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/m68k: Merge disas_m68k_insn into m68k_tr_translate_insn

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20180512050250.12774-10-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

target/m68k: Improve ending TB at page boundaries

Rather than limit total TB size to PAGE-32 bytes, end the TB when
near the end of a page.  This should provide proper semantics of
SIGSEGV when executing near the end of a page.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20180512050250.12774-9-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

target/m68k: Convert to TranslatorOps

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20180512050250.12774-8-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

target/m68k: Convert to DisasContextBase

Removed ctx->insn_pc in favour of ctx->base.pc_next.
Yes, it is annoying, but didn't want to waste its 4 bytes.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20180512050250.12774-7-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

Merge remote-tracking branch 'remotes/jnsnow/tags/ide-pull-request' into staging

Pull request

# gpg: Signature made Fri 08 Jun 2018 18:46:24 BST
# gpg:                using RSA key 7DEF8106AAFC390E
# gpg: Good signature from "John Snow (John Huston) <jsnow@redhat.com>"
# Primary key fingerprint: FAEB 9711 A12C F475 812F  18F2 88A9 064D 1835 61EB
#      Subkey fingerprint: F9B7 ABDB BCAC DF95 BE76  CBD0 7DEF 8106 AAFC 390E

* remotes/jnsnow/tags/ide-pull-request: (30 commits)
  ide: introduce ide_transfer_start_norecurse
  atapi: call ide_set_irq before ide_transfer_start
  ide: make ide_transfer_stop idempotent
  ide: call ide_cmd_done from ide_transfer_stop
  ide: push end_transfer_func out of start_transfer callback, rename callback
  ahci: move PIO Setup FIS before transfer, fix it for ATAPI commands
  libqos/ahci: track sector size
  MAINTAINERS: Add the cdrom-test to John's section
  tests/cdrom-test: Test that -cdrom parameter is working
  tests/cdrom-test: Test booting from CD-ROM ISO image file
  tests/boot-sector: Add magic bytes to s390x boot code header
  ahci: make ahci_mem_write traces more descriptive
  ahci: delete old host register address definitions
  ahci: adjust ahci_mem_write to work on registers
  ahci: fix spacing damage on ahci_mem_write
  ahci: make mem_read_32 traces more descriptive
  ahci: modify ahci_mem_read_32 to work on register numbers
  ahci: fix host register max address
  ahci: add host register enumeration
  ahci: delete old port register address definitions
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/m68k: Rename DISAS_UPDATE and gen_lookup_tb

The name gen_lookup_tb is at odds with tcg_gen_lookup_and_goto_tb.
For these cases, we do indeed want to exit back to the main loop.
Similarly, DISAS_UPDATE performs no actual update, whereas DISAS_EXIT
does what it says.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20180512050250.12774-6-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

target/m68k: Use lookup_and_goto_tb for DISAS_JUMP

These are all indirect or out-of-page direct jumps.
We can indirectly chain to the next TB without going
back to the main loop.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20180512050250.12774-5-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

target/m68k: Remove DISAS_JUMP_NEXT as unused

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20180512050250.12774-4-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

target/m68k: Replace DISAS_TB_JUMP with DISAS_NORETURN

We have exited the TB after using goto_tb; there is no
distinction from DISAS_NORETURN.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20180512050250.12774-3-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

target/m68k: Use DISAS_NORETURN for exceptions

The raise_exception helper does not return.  Do not generate
any code following that.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20180512050250.12774-2-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

ide: introduce ide_transfer_start_norecurse

For the case where the end_transfer_func is also the caller of
ide_transfer_start, the mutual recursion can lead to unlimited
stack usage.  Introduce a new version that can be used to change
tail recursion into a loop, and use it in trace_ide_atapi_cmd_reply_end.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180606190955.20845-8-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

atapi: call ide_set_irq before ide_transfer_start

The ATAPI_INT_REASON_IO interrupt is raised when I/O starts, but in the
AHCI case ide_set_irq was actually called at the end of a mutual recursion.
Move it early, with the side effect that ide_transfer_start becomes a tail
call in ide_atapi_cmd_reply_end.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180606190955.20845-7-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ide: make ide_transfer_stop idempotent

There is code checking s->end_transfer_func and it was not taught about
ide_transfer_cancel.  We can just use ide_transfer_stop because
s->end_transfer_func is only ever called in the DRQ phase.

ide_transfer_cancel can then be removed, since it would just be
calling ide_transfer_halt.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180606190955.20845-6-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ide: call ide_cmd_done from ide_transfer_stop

The code can simply be moved to the sole caller that has notify == true.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180606190955.20845-5-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ide: push end_transfer_func out of start_transfer callback, rename callback

Now that end_transfer_func is a tail call in ahci_start_transfer,
formalize the fact that the callback (of which ahci_start_transfer is
the sole implementation) takes care of the transfer too: rename it to
pio_transfer and, if it is present, call the end_transfer_func as soon
as it returns.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180606190955.20845-4-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: move PIO Setup FIS before transfer, fix it for ATAPI commands

The PIO Setup FIS is written in the PIO:Entry state, which comes before
the ATA and ATAPI data transfer states.  As a result, the PIO Setup FIS
interrupt is now raised before DMA ends for ATAPI commands, and tests have
to be adjusted.

This is also hinted by the description of the command header in the AHCI
specification, where the "A" bit is described as

    When ‘1’, indicates that a PIO setup FIS shall be sent by the device
    indicating a transfer for the ATAPI command.

and also by the description of the ACMD (ATAPI command region):

    The ATAPI command must be either 12 or 16 bytes in length. The length
    transmitted by the HBA is determined by the PIO setup FIS that is sent
    by the device requesting the ATAPI command.

QEMU, which conflates the "generator" and the "receiver" of the FIS into
one device, always uses ATAPI_PACKET_SIZE, aka 12, for the length.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180606190955.20845-3-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

libqos/ahci: track sector size

It's not always 512, and it does wind up mattering for PIO tranfers,
because this means DRQ blocks are four times as big for ATAPI.
Replace an instance of 2048 with the correct define, too.

This patch by itself winds changing no behavior. fis->count is ignored
for CMD_PACKET, and sect_count only gets used in non-ATAPI cases.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180606190955.20845-2-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

MAINTAINERS: Add the cdrom-test to John's section

The cdrom-test checks various block types - IDE, SCSI and
virtio, so it's a little bit hard to decide where this should
belong to in the MAINTAINERS file. But John volunteered to take
it, so let's put it into the IDE section for now.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Acked-by: John Snow <jsnow@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>

tests/cdrom-test: Test that -cdrom parameter is working

Commit 1454509726719e0933c800 recently broke the "-cdrom" parameter
on a couple of boards without us noticing it immediately. Thus let's
add a test which checks that "-cdrom" can at least be used to start
QEMU with certain machine types.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Hervé Poussineau <hpoussin@reactos.org>
Acked-By: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>

tests/cdrom-test: Test booting from CD-ROM ISO image file

We already have the code for a boot file in tests/boot-sector.c,
so if the genisoimage program is available, we can easily create
a bootable CD ISO image that we can use for testing whether our
CD-ROM emulation and the BIOS CD-ROM boot works correctly.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Hervé Poussineau <hpoussin@reactos.org>
Acked-By: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>

tests/boot-sector: Add magic bytes to s390x boot code header

We're going to use the s390x boot code for testing CD-ROM booting.
But the ISO loader of the s390-ccw bios is a little bit more picky
than the network loader and expects some magic bytes in the header
of the file (see linux_s390_magic in pc-bios/s390-ccw/bootmap.c), so
we've got to add them in our boot code here, too.

Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Hervé Poussineau <hpoussin@reactos.org>
Acked-By: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: make ahci_mem_write traces more descriptive

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-17-jsnow@redhat.com
[Fixed format specifiers. --js]
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: delete old host register address definitions

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-16-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: adjust ahci_mem_write to work on registers

Actually, this function looks pretty broken, but for now, let's finish
up what this series of commits came here to do.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-15-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: fix spacing damage on ahci_mem_write

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-14-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: make mem_read_32 traces more descriptive

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-13-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: modify ahci_mem_read_32 to work on register numbers

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-12-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: fix host register max address

Yes, comment, it ought to be 0x2C.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-11-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: add host register enumeration

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-10-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: delete old port register address definitions

They're now unused.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-9-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: make port write traces more descriptive

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-8-jsnow@redhat.com
[Changed format specifier. --js]
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: modify ahci_port_write to use register numbers

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-7-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: combine identical clauses in port write

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-6-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: fix spacing damage on ahci_port_write

Churn.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-5-jsnow@redhat.com
[Fix patchew/checkpatch nit. --js]
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: make port read traces more descriptive

A trace is added to let us watch unimplemented registers specifically,
as these are more likely to cause us trouble. Otherwise, the port read
traces now tell us what register is getting hit, which is nicer.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-4-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: modify ahci_port_read to use register numbers

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-3-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: add port register enumeration

Instead of tracking offsets, lets count the registers.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20180531222835.16558-2-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: don't schedule unnecessary BH

The comment gives us a hint. *Maybe* we still have something to
process. Well, why not check?

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Message-id: 20180531004323.4611-4-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>

ahci: fix PxCI register race

Fixes: https://bugs.launchpad.net/qemu/+bug/1769189
AHCI presently signals completion prior to the PxCI register being
cleared to indicate completion. If a guest driver attempts to issue
a new command in its IRQ handler, it might be surprised to learn there
is still a command pending.

In the case of Windows 10's boot driver, it will actually poll the IRQ
register hoping to find out when the command is done running -- which
will never happen, as there isn't a command running.

Fix this: clear PxCI in ahci_cmd_done and not in the asynchronous BH.
Because it now runs synchronously, we don't need to check if the command
is actually done by spying on the ATA registers. We know it's done.

CC: qemu-stable <qemu-stable@nongnu.org>
Reported-by: François Guerraz <kubrick@fgv6.net>
Tested-by: Bruce Rogers <brogers@suse.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Message-id: 20180531004323.4611-3-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>