habanalabs: allow user to set allocation page size

In future ASICs the MMU will be able to work with multiple page sizes,
thus a new flag is added to allow the user to set the requested page
size.

This flag is added since the whole DRAM is allocated for the user and
the user also should be familiar with the memory usage use case.

As such, the user may choose to "over allocate" memory in favor of
performance (for instance- large page allocations covers more memory
in less TLB entries).

For example: say available page sizes are of 1MB and 32MB. If user
wants to allocate 40MB the user can either set page size to 1MB and
allocate the exact amount of memory (but will result in 40 TLB entries)
or the user can use 32MB pages, "waste" 8MB of physical memory but
occupy only 2 TLB entries.

Note that this feature will be available only to ASIC that supports
multiple DRAM page sizes.

Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index b5055ab528b89bb2323ee98df53acf55faf408e0..677ae4ff922cafe441173a73d26c669702963e24 100644 (file)
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -563,6 +563,7 @@ struct hl_hints_range {
  *                              devices)
  * @configurable_stop_on_err: is stop-on-error option configurable via debugfs.
  * @set_max_power_on_device_init: true if need to set max power in F/W on device init.
+ * @supports_user_set_page_size: true if user can set the allocation page size.
  */
 struct asic_fixed_properties {
        struct hw_queue_properties      *hw_queues_props;
@@ -648,6 +649,7 @@ struct asic_fixed_properties {
        u8                              allow_inference_soft_reset;
        u8                              configurable_stop_on_err;
        u8                              set_max_power_on_device_init;
+       u8                              supports_user_set_page_size;
 };
 
 /**
@@ -1311,6 +1313,8 @@ struct fw_load_mgr {
  * @get_sob_addr: get SOB base address offset.
  * @set_pci_memory_regions: setting properties of PCI memory regions
  * @get_stream_master_qid_arr: get pointer to stream masters QID array
+ * @is_valid_dram_page_size: return true if page size is supported in device
+ *                           memory allocation, otherwise false.
  */
 struct hl_asic_funcs {
        int (*early_init)(struct hl_device *hdev);
@@ -1436,6 +1440,7 @@ struct hl_asic_funcs {
        u32 (*get_sob_addr)(struct hl_device *hdev, u32 sob_id);
        void (*set_pci_memory_regions)(struct hl_device *hdev);
        u32* (*get_stream_master_qid_arr)(void);
+       bool (*is_valid_dram_page_size)(u32 page_size);
 };
 
 

diff --git a/drivers/misc/habanalabs/common/memory.c b/drivers/misc/habanalabs/common/memory.c
index 7bd4cae87a549e479eafd4b9a70b636aad4a243e..385bf3448c730b84964250162d4f8e8cb79363a9 100644 (file)
--- a/drivers/misc/habanalabs/common/memory.c
+++ b/drivers/misc/habanalabs/common/memory.c
@@ -24,6 +24,31 @@ MODULE_IMPORT_NS(DMA_BUF);
 static int allocate_timestamps_buffers(struct hl_fpriv *hpriv,
                        struct hl_mem_in *args, u64 *handle);
 
+static int set_alloc_page_size(struct hl_device *hdev, struct hl_mem_in *args, u32 *page_size)
+{
+       struct asic_fixed_properties *prop = &hdev->asic_prop;
+       u32 psize;
+
+       /*
+        * for ASIC that supports setting the allocation page size by user we will address
+        * user's choice only if it is not 0 (as 0 means taking the default page size)
+        */
+       if (prop->supports_user_set_page_size && args->alloc.page_size) {
+               psize = args->alloc.page_size;
+
+               if (!hdev->asic_funcs->is_valid_dram_page_size(psize)) {
+                       dev_err(hdev->dev, "user page size (%#x) is not valid\n", psize);
+                       return -EINVAL;
+               }
+       } else {
+               psize = hdev->asic_prop.dram_page_size;
+       }
+
+       *page_size = psize;
+
+       return 0;
+}
+
 /*
  * The va ranges in context object contain a list with the available chunks of
  * device virtual memory.
@@ -69,7 +94,11 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
        bool contiguous;
 
        num_curr_pgs = 0;
-       page_size = hdev->asic_prop.dram_page_size;
+
+       rc = set_alloc_page_size(hdev, args, &page_size);
+       if (rc)
+               return rc;
+
        num_pgs = DIV_ROUND_UP_ULL(args->alloc.mem_size, page_size);
        total_size = num_pgs * page_size;
 

diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index ad640a268a3c8d8c19a133c5501ad94ac7d33934..a60c0e23d23c52054988bd0ae7e59def4837bf26 100644 (file)
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -9432,7 +9432,8 @@ static const struct hl_asic_funcs gaudi_funcs = {
        .state_dump_init = gaudi_state_dump_init,
        .get_sob_addr = gaudi_get_sob_addr,
        .set_pci_memory_regions = gaudi_set_pci_memory_regions,
-       .get_stream_master_qid_arr = gaudi_get_stream_master_qid_arr
+       .get_stream_master_qid_arr = gaudi_get_stream_master_qid_arr,
+       .is_valid_dram_page_size = NULL
 };
 
 /**

diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index c08d96e43c9f40250233987d2448b0195af0e025..ec9358bcbf0bdf0efed92e5f0d64a043f5b1d6b8 100644 (file)
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -5764,6 +5764,7 @@ static const struct hl_asic_funcs goya_funcs = {
        .get_sob_addr = &goya_get_sob_addr,
        .set_pci_memory_regions = goya_set_pci_memory_regions,
        .get_stream_master_qid_arr = goya_get_stream_master_qid_arr,
+       .is_valid_dram_page_size = NULL
 };
 
 /*

diff --git a/include/uapi/misc/habanalabs.h b/include/uapi/misc/habanalabs.h
index a2a953a91b5e24adf043ec03b39866b3adfc577f..1d6b4f0c41594fc6774d46b585609dcc1e69df66 100644 (file)
--- a/include/uapi/misc/habanalabs.h
+++ b/include/uapi/misc/habanalabs.h
@@ -1,6 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note
  *
- * Copyright 2016-2021 HabanaLabs, Ltd.
+ * Copyright 2016-2022 HabanaLabs, Ltd.
  * All Rights Reserved.
  *
  */
@@ -1130,9 +1130,12 @@ struct hl_mem_in {
                /**
                 * structure for device memory allocation (used with the HL_MEM_OP_ALLOC op)
                 * @mem_size: memory size to allocate
+                * @page_size: page size to use on allocation. when the value is 0 the default page
+                *             size will be taken.
                 */
                struct {
                        __u64 mem_size;
+                       __u64 page_size;
                } alloc;
 
                /**