* [PATCH v2 1/3] drm/amdgpu: unwind TTM on amdgpu_ttm_init() failure
2026-11-18 19:02 [PATCH v2 0/3] fix TTM device_list use-after-free on failed probe Mario Limonciello
@ 2026-11-18 19:02 ` Mario Limonciello
2026-11-18 19:03 ` [PATCH v2 2/3] drm/amdgpu: unwind bo_init on GMC sw_init failure Mario Limonciello
2026-11-18 19:03 ` [PATCH v2 3/3] drm/amdgpu: release WC memtype and MTRR when amdgpu_bo_init() fails Mario Limonciello
2 siblings, 0 replies; 5+ messages in thread
From: Mario Limonciello @ 2026-11-18 19:02 UTC (permalink / raw)
To: Alex Deucher; +Cc: Mario Limonciello, amd-gfx
If amdgpu_ttm_init() fails after ttm_device_init() succeeded, the error
paths just return. Nothing later calls ttm_device_fini() because the GMC
IP block never reaches sw_fini, so the freed amdgpu_device stays linked
on ttm_glob.device_list. The next ttm_device_init() or
ttm_global_swapout() then walks freed memory. The MMIO_REMAP BO created
in amdgpu_ttm_init() is also leaked, which shows up as "Memory manager
not clean" on the following bind.
Route every error after ttm_device_init() through amdgpu_ttm_fini(), and
mark the manager initialized as soon as ttm_device_init() succeeds so the
unwind also covers a pool init failure. Make amdgpu_ttm_fini() skip the
VRAM, GTT and PREEMPT managers when they were never registered, as it can
now run on a partially initialized device.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 43 ++++++++++++++++++++-------------
1 file changed, 26 insertions(+), 17 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 8dfa92967b630..7828caf8aba0d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -2153,20 +2153,21 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
"failed initializing buffer object driver(%d).\n", r);
return r;
}
+ /* from here on amdgpu_ttm_fini() is responsible for unwinding */
+ adev->mman.initialized = true;
r = amdgpu_ttm_pools_init(adev);
if (r) {
dev_err(adev->dev, "failed to init ttm pools(%d).\n", r);
- return r;
+ goto err_fini;
}
- adev->mman.initialized = true;
if (!adev->gmc.is_app_apu) {
/* Initialize VRAM pool with all of VRAM divided into pages */
r = amdgpu_vram_mgr_init(adev);
if (r) {
dev_err(adev->dev, "Failed initializing VRAM heap.\n");
- return r;
+ goto err_fini;
}
}
@@ -2190,7 +2191,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
r = amdgpu_ttm_alloc_vram_resv_regions(adev);
if (r)
- return r;
+ goto err_fini;
/* RAS loads and reserves bad pages */
amdgpu_ras_mgr_early_init_service(adev);
@@ -2239,7 +2240,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
r = amdgpu_gtt_mgr_init(adev, gtt_size);
if (r) {
dev_err(adev->dev, "Failed initializing GTT heap.\n");
- return r;
+ goto err_fini;
}
dev_info(adev->dev, " %uM of GTT memory ready.\n",
(unsigned int)(gtt_size / (1024 * 1024)));
@@ -2253,52 +2254,52 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_DOORBELL, adev->doorbell.size / PAGE_SIZE);
if (r) {
dev_err(adev->dev, "Failed initializing doorbell heap.\n");
- return r;
+ goto err_fini;
}
/* Create a doorbell page for kernel usages */
r = amdgpu_doorbell_create_kernel_doorbells(adev);
if (r) {
dev_err(adev->dev, "Failed to initialize kernel doorbells.\n");
- return r;
+ goto err_fini;
}
/* Initialize MMIO-remap pool (single page 4K) */
r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_MMIO_REMAP, 1);
if (r) {
dev_err(adev->dev, "Failed initializing MMIO-remap heap.\n");
- return r;
+ goto err_fini;
}
/* Allocate the singleton MMIO_REMAP BO if supported */
r = amdgpu_ttm_alloc_mmio_remap_bo(adev);
if (r)
- return r;
+ goto err_fini;
/* Initialize preemptible memory pool */
r = amdgpu_preempt_mgr_init(adev);
if (r) {
dev_err(adev->dev, "Failed initializing PREEMPT heap.\n");
- return r;
+ goto err_fini;
}
/* Initialize various on-chip memory pools */
r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_GDS, adev->gds.gds_size);
if (r) {
dev_err(adev->dev, "Failed initializing GDS heap.\n");
- return r;
+ goto err_fini;
}
r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_GWS, adev->gds.gws_size);
if (r) {
dev_err(adev->dev, "Failed initializing gws heap.\n");
- return r;
+ goto err_fini;
}
r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_OA, adev->gds.oa_size);
if (r) {
dev_err(adev->dev, "Failed initializing oa heap.\n");
- return r;
+ goto err_fini;
}
if (adev->ualink.npa_size) {
@@ -2306,7 +2307,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
adev->ualink.npa_size);
if (r) {
dev_err(adev->dev, "Failed initializing NPA heap.\n");
- return r;
+ goto err_fini;
}
}
@@ -2318,6 +2319,10 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
"Debug VRAM access will use slowpath MM access\n");
return 0;
+
+err_fini:
+ amdgpu_ttm_fini(adev);
+ return r;
}
/*
@@ -2352,10 +2357,14 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev)
adev->mman.aper_base_kaddr = NULL;
}
- if (!adev->gmc.is_app_apu)
+ /* may be called on a partially initialized TTM, skip unregistered managers */
+ if (!adev->gmc.is_app_apu &&
+ ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM))
amdgpu_vram_mgr_fini(adev);
- amdgpu_gtt_mgr_fini(adev);
- amdgpu_preempt_mgr_fini(adev);
+ if (ttm_manager_type(&adev->mman.bdev, TTM_PL_TT))
+ amdgpu_gtt_mgr_fini(adev);
+ if (ttm_manager_type(&adev->mman.bdev, AMDGPU_PL_PREEMPT))
+ amdgpu_preempt_mgr_fini(adev);
amdgpu_doorbell_fini(adev);
ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_GDS);
--
2.43.0
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH v2 2/3] drm/amdgpu: unwind bo_init on GMC sw_init failure
2026-11-18 19:02 [PATCH v2 0/3] fix TTM device_list use-after-free on failed probe Mario Limonciello
2026-11-18 19:02 ` [PATCH v2 1/3] drm/amdgpu: unwind TTM on amdgpu_ttm_init() failure Mario Limonciello
@ 2026-11-18 19:03 ` Mario Limonciello
2026-11-18 19:03 ` [PATCH v2 3/3] drm/amdgpu: release WC memtype and MTRR when amdgpu_bo_init() fails Mario Limonciello
2 siblings, 0 replies; 5+ messages in thread
From: Mario Limonciello @ 2026-11-18 19:03 UTC (permalink / raw)
To: Alex Deucher; +Cc: Mario Limonciello, amd-gfx
When gmc_v*_sw_init() fails after amdgpu_bo_init(), the GMC block's
status.sw stays false, so amdgpu_device_ip_fini() skips its sw_fini and
amdgpu_bo_fini() (and thus ttm_device_fini()) is never called. The
freed device is left on ttm_glob.device_list.
Unwind on the failure paths after amdgpu_bo_init(): GART init failures
call amdgpu_bo_fini(); later failures (RAS sw init on v9-v12,
vm_fault_info allocation on v7/v8) also call amdgpu_vm_manager_fini().
The GART table BO (and pdb0_bo on v9/v12) is pinned in VRAM by
gmc_v*_gart_init(), and pinned BOs can't be evicted during TTM teardown.
Free them before amdgpu_bo_fini(), as gmc_v*_sw_fini() does, otherwise
ttm_device_fini() warns and leaks the BOs and their backing allocations.
On v9 the GART table lives in system memory when there is no VRAM, so
use the matching RAM free helper there, guarded against gart_init()
failing before the BO was created.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 11 +++++++++--
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 11 +++++++++--
drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c | 12 ++++++++++--
drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c | 7 ++++++-
drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c | 15 ++++++++++++---
drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 15 ++++++++++++---
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 17 +++++++++++++++--
7 files changed, 73 insertions(+), 15 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index 49f593fe2a876..61e4b535d0a30 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -861,7 +861,7 @@ static int gmc_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
r = gmc_v10_0_gart_init(adev);
if (r)
- return r;
+ goto err_bo_fini;
/*
* number of VMs
@@ -880,9 +880,16 @@ static int gmc_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
r = amdgpu_gmc_ras_sw_init(adev);
if (r)
- return r;
+ goto err_vm_fini;
return 0;
+
+err_vm_fini:
+ amdgpu_vm_manager_fini(adev);
+err_bo_fini:
+ amdgpu_gart_table_vram_free(adev);
+ amdgpu_bo_fini(adev);
+ return r;
}
/**
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
index bc3b1ffb54886..3879e5a817e42 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
@@ -859,7 +859,7 @@ static int gmc_v11_0_sw_init(struct amdgpu_ip_block *ip_block)
r = gmc_v11_0_gart_init(adev);
if (r)
- return r;
+ goto err_bo_fini;
/*
* number of VMs
@@ -879,9 +879,16 @@ static int gmc_v11_0_sw_init(struct amdgpu_ip_block *ip_block)
r = amdgpu_gmc_ras_sw_init(adev);
if (r)
- return r;
+ goto err_vm_fini;
return 0;
+
+err_vm_fini:
+ amdgpu_vm_manager_fini(adev);
+err_bo_fini:
+ amdgpu_gart_table_vram_free(adev);
+ amdgpu_bo_fini(adev);
+ return r;
}
/**
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c
index 3bb13cfed92fe..b4ed5783ed5fb 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c
@@ -987,7 +987,7 @@ static int gmc_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
r = gmc_v12_0_gart_init(adev);
if (r)
- return r;
+ goto err_bo_fini;
if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 1, 0))
gmc_v12_1_init_nps_details(adev);
@@ -1021,12 +1021,20 @@ static int gmc_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
r = amdgpu_gmc_ras_sw_init(adev);
if (r)
- return r;
+ goto err_vm_fini;
if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 1, 0))
amdgpu_gmc_sysfs_init(adev);
return 0;
+
+err_vm_fini:
+ amdgpu_vm_manager_fini(adev);
+err_bo_fini:
+ amdgpu_gart_table_vram_free(adev);
+ amdgpu_bo_free_kernel(&adev->gmc.pdb0_bo, NULL, &adev->gmc.ptr_pdb0);
+ amdgpu_bo_fini(adev);
+ return r;
}
/**
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
index a7c353d1ac09d..135824824e53d 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
@@ -817,7 +817,7 @@ static int gmc_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
r = gmc_v6_0_gart_init(adev);
if (r)
- return r;
+ goto err_bo_fini;
/*
* number of VMs
@@ -840,6 +840,11 @@ static int gmc_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
}
return 0;
+
+err_bo_fini:
+ amdgpu_gart_table_vram_free(adev);
+ amdgpu_bo_fini(adev);
+ return r;
}
static int gmc_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
index c09f5153679ea..565054d14d6db 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
@@ -999,7 +999,7 @@ static int gmc_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
r = gmc_v7_0_gart_init(adev);
if (r)
- return r;
+ goto err_bo_fini;
/*
* number of VMs
@@ -1026,11 +1026,20 @@ static int gmc_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
}
adev->gmc.vm_fault_info = kmalloc_obj(struct kfd_vm_fault_info);
- if (!adev->gmc.vm_fault_info)
- return -ENOMEM;
+ if (!adev->gmc.vm_fault_info) {
+ r = -ENOMEM;
+ goto err_vm_fini;
+ }
atomic_set_release(&adev->gmc.vm_fault_info_updated, 0);
return 0;
+
+err_vm_fini:
+ amdgpu_vm_manager_fini(adev);
+err_bo_fini:
+ amdgpu_gart_table_vram_free(adev);
+ amdgpu_bo_fini(adev);
+ return r;
}
static int gmc_v7_0_sw_fini(struct amdgpu_ip_block *ip_block)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
index 15592027e7c0c..986ff114f06a1 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
@@ -1114,7 +1114,7 @@ static int gmc_v8_0_sw_init(struct amdgpu_ip_block *ip_block)
r = gmc_v8_0_gart_init(adev);
if (r)
- return r;
+ goto err_bo_fini;
/*
* number of VMs
@@ -1141,11 +1141,20 @@ static int gmc_v8_0_sw_init(struct amdgpu_ip_block *ip_block)
}
adev->gmc.vm_fault_info = kmalloc_obj(struct kfd_vm_fault_info);
- if (!adev->gmc.vm_fault_info)
- return -ENOMEM;
+ if (!adev->gmc.vm_fault_info) {
+ r = -ENOMEM;
+ goto err_vm_fini;
+ }
atomic_set_release(&adev->gmc.vm_fault_info_updated, 0);
return 0;
+
+err_vm_fini:
+ amdgpu_vm_manager_fini(adev);
+err_bo_fini:
+ amdgpu_gart_table_vram_free(adev);
+ amdgpu_bo_fini(adev);
+ return r;
}
static int gmc_v8_0_sw_fini(struct amdgpu_ip_block *ip_block)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index a705830ef436c..5d40e2020b4a8 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -2026,7 +2026,7 @@ static int gmc_v9_0_sw_init(struct amdgpu_ip_block *ip_block)
r = gmc_v9_0_gart_init(adev);
if (r)
- return r;
+ goto err_bo_fini;
gmc_v9_0_init_nps_details(adev);
/*
@@ -2066,12 +2066,25 @@ static int gmc_v9_0_sw_init(struct amdgpu_ip_block *ip_block)
r = amdgpu_gmc_ras_sw_init(adev);
if (r)
- return r;
+ goto err_vm_fini;
if (amdgpu_is_multi_aid(adev))
amdgpu_gmc_sysfs_init(adev);
return 0;
+
+err_vm_fini:
+ amdgpu_vm_manager_fini(adev);
+err_bo_fini:
+ if (!adev->gmc.real_vram_size) {
+ /* GART bo may not exist if gart_init failed early */
+ if (adev->gart.bo)
+ amdgpu_gart_table_ram_free(adev);
+ } else
+ amdgpu_gart_table_vram_free(adev);
+ amdgpu_bo_free_kernel(&adev->gmc.pdb0_bo, NULL, &adev->gmc.ptr_pdb0);
+ amdgpu_bo_fini(adev);
+ return r;
}
static int gmc_v9_0_sw_fini(struct amdgpu_ip_block *ip_block)
--
2.43.0
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH v2 3/3] drm/amdgpu: release WC memtype and MTRR when amdgpu_bo_init() fails
2026-11-18 19:02 [PATCH v2 0/3] fix TTM device_list use-after-free on failed probe Mario Limonciello
2026-11-18 19:02 ` [PATCH v2 1/3] drm/amdgpu: unwind TTM on amdgpu_ttm_init() failure Mario Limonciello
2026-11-18 19:03 ` [PATCH v2 2/3] drm/amdgpu: unwind bo_init on GMC sw_init failure Mario Limonciello
@ 2026-11-18 19:03 ` Mario Limonciello
2026-11-18 19:09 ` sashiko-bot
2 siblings, 1 reply; 5+ messages in thread
From: Mario Limonciello @ 2026-11-18 19:03 UTC (permalink / raw)
To: Alex Deucher; +Cc: Mario Limonciello, amd-gfx
amdgpu_bo_init() reserves the WC memtype for the VRAM aperture and adds
an MTRR before calling amdgpu_ttm_init(). If amdgpu_ttm_init() fails,
the GMC sw_init callers just return the error and GMC's sw_fini is never
run, so amdgpu_bo_fini(), the only place that drops the memtype and MTRR,
is never called. A later rebind then reserves the same aperture again
with the old reservation still held.
Undo the memtype reservation and the MTRR in amdgpu_bo_init() itself when
amdgpu_ttm_init() fails.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 20 +++++++++++++++++---
1 file changed, 17 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index d4651f2d437c6..e28d7619187af 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -1130,11 +1130,13 @@ static const char * const amdgpu_vram_names[] = {
*/
int amdgpu_bo_init(struct amdgpu_device *adev)
{
+ int r;
+
/* On A+A platform, VRAM can be mapped as WB */
if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
/* reserve PAT memory space to WC for VRAM */
- int r = arch_io_reserve_memtype_wc(adev->gmc.aper_base,
- adev->gmc.aper_size);
+ r = arch_io_reserve_memtype_wc(adev->gmc.aper_base,
+ adev->gmc.aper_size);
if (r) {
DRM_ERROR("Unable to set WC memtype for the aperture base\n");
@@ -1151,7 +1153,19 @@ int amdgpu_bo_init(struct amdgpu_device *adev)
(unsigned long long)adev->gmc.aper_size >> 20);
drm_info(adev_to_drm(adev), "RAM width %dbits %s\n",
adev->gmc.vram_width, amdgpu_vram_names[adev->gmc.vram_type]);
- return amdgpu_ttm_init(adev);
+
+ r = amdgpu_ttm_init(adev);
+ if (r) {
+ /* the caller does not call amdgpu_bo_fini() when we fail */
+ if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
+ arch_phys_wc_del(adev->gmc.vram_mtrr);
+ arch_io_free_memtype_wc(adev->gmc.aper_base,
+ adev->gmc.aper_size);
+ }
+ return r;
+ }
+
+ return 0;
}
/**
--
2.43.0
^ permalink raw reply related [flat|nested] 5+ messages in thread* Re: [PATCH v2 3/3] drm/amdgpu: release WC memtype and MTRR when amdgpu_bo_init() fails
2026-11-18 19:03 ` [PATCH v2 3/3] drm/amdgpu: release WC memtype and MTRR when amdgpu_bo_init() fails Mario Limonciello
@ 2026-11-18 19:09 ` sashiko-bot
0 siblings, 0 replies; 5+ messages in thread
From: sashiko-bot @ 2026-11-18 19:09 UTC (permalink / raw)
To: Mario Limonciello; +Cc: amd-gfx
Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Low] drm/amdgpu: missing Fixes trailer for amdgpu_bo_init() leak fix
--
commit b39bca7df25d1f904ab6219bf84105076d42ff05
Author: Mario Limonciello <mario.limonciello@amd.com>
> drm/amdgpu: release WC memtype and MTRR when amdgpu_bo_init() fails
>
> amdgpu_bo_init() reserves the WC memtype for the VRAM aperture and adds
> an MTRR before calling amdgpu_ttm_init(). If amdgpu_ttm_init() fails,
> the GMC sw_init callers just return the error and GMC's sw_fini is never
> run, so amdgpu_bo_fini(), the only place that drops the memtype and MTRR,
> is never called. A later rebind then reserves the same aperture again
> with the old reservation still held.
>
> Undo the memtype reservation and the MTRR in amdgpu_bo_init() itself when
> amdgpu_ttm_init() fails.
>
> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
[Severity: Low]
Should this bug fix include Fixes tags for the commits that introduced the
MTRR and WC memtype reservations in amdgpu_bo_init() without unwinding
them on failure?
Fixes: d38ceaf99ed0 ("drm/amdgpu: add core driver (v4)")
Fixes: 7cf321d118a8 ("drm/drivers: add support for using the arch wc mapping API.")
--
Sashiko AI review · https://sashiko.dev/#/patchset/20261118-superm1-amdgpu-memory-leak-v2-0-38365f37bd5a@amd.com?part=3
^ permalink raw reply [flat|nested] 5+ messages in thread