diff mbox series

[RFC,1/2] drm/amdgpu: Fix memory corruption due to swapout and swapin

Message ID 20210519022852.16766-1-xinhui.pan@amd.com (mailing list archive)
State New, archived
Headers show
Series [RFC,1/2] drm/amdgpu: Fix memory corruption due to swapout and swapin | expand

Commit Message

Pan, Xinhui May 19, 2021, 2:28 a.m. UTC
cpu 1                                           cpu 2
kfd alloc BO A(userptr)                         alloc BO B(GTT)
    ->init -> validate				-> init -> validate -> populate
    init_user_pages				  -> swapout BO A //hit ttm pages limit
	-> get_user_pages (fill up ttm->pages)
	 -> validate -> populate
          -> swapin BO A // Now hit the BUG

We know that get_user_pages may race with swapout on same BO.
Threre are some issues I have met.
1) memory corruption.
This is because we do a swap before memory is setup. ttm_tt_swapout()
just create a swap_storage with its content being 0x0. So when we setup
memory after the swapout. The following swapin makes the memory
corrupted.

2) panic
When swapout happes with get_user_pages, they touch ttm->pages without
anylock. It causes memory corruption too. But I hit page fault mostly.

Signed-off-by: xinhui pan <xinhui.pan@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

Comments

Felix Kuehling May 19, 2021, 3:29 a.m. UTC | #1
Swapping SG BOs makes no sense, because TTM doesn't own the pages of 
this type of BO.

Last I checked, userptr BOs (and other SG BOs) were protected from 
swapout by the fact that they would not be added to the swap-LRU. But it 
looks like Christian just removed the swap-LRU. I guess this broke that 
protection:

commit 2cb51d22d70b18eaf339abf9758bf0b7608da65c
Author: Christian König <christian.koenig@amd.com>
Date:   Tue Oct 6 16:30:09 2020 +0200

     drm/ttm: remove swap LRU v3

     Instead evict round robin from each devices SYSTEM and TT domain.

     v2: reorder num_pages access reported by Dan's script
     v3: fix rebase fallout, num_pages should be 32bit

     Signed-off-by: Christian König <christian.koenig@amd.com>
     Tested-by: Nirmoy Das <nirmoy.das@amd.com>
     Reviewed-by: Huang Rui <ray.huang@amd.com>
     Reviewed-by: Matthew Auld <matthew.auld@intel.com>
     Link: https://patchwork.freedesktop.org/patch/424009/

Regards,
   Felix


On 2021-05-18 10:28 p.m., xinhui pan wrote:
> cpu 1                                           cpu 2
> kfd alloc BO A(userptr)                         alloc BO B(GTT)
>      ->init -> validate				-> init -> validate -> populate
>      init_user_pages				  -> swapout BO A //hit ttm pages limit
> 	-> get_user_pages (fill up ttm->pages)
> 	 -> validate -> populate
>            -> swapin BO A // Now hit the BUG
>
> We know that get_user_pages may race with swapout on same BO.
> Threre are some issues I have met.
> 1) memory corruption.
> This is because we do a swap before memory is setup. ttm_tt_swapout()
> just create a swap_storage with its content being 0x0. So when we setup
> memory after the swapout. The following swapin makes the memory
> corrupted.
>
> 2) panic
> When swapout happes with get_user_pages, they touch ttm->pages without
> anylock. It causes memory corruption too. But I hit page fault mostly.
>
> Signed-off-by: xinhui pan <xinhui.pan@amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 16 +++++++++++++++-
>   1 file changed, 15 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> index 928e8d57cd08..42460e4480f8 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> @@ -835,6 +835,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>   	struct amdkfd_process_info *process_info = mem->process_info;
>   	struct amdgpu_bo *bo = mem->bo;
>   	struct ttm_operation_ctx ctx = { true, false };
> +	struct page **pages;
>   	int ret = 0;
>   
>   	mutex_lock(&process_info->lock);
> @@ -852,7 +853,13 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>   		goto out;
>   	}
>   
> -	ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages);
> +	pages = kvmalloc_array(bo->tbo.ttm->num_pages,
> +			sizeof(struct page *),
> +			GFP_KERNEL | __GFP_ZERO);
> +	if (!pages)
> +		goto unregister_out;
> +
> +	ret = amdgpu_ttm_tt_get_user_pages(bo, pages);
>   	if (ret) {
>   		pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
>   		goto unregister_out;
> @@ -863,6 +870,12 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>   		pr_err("%s: Failed to reserve BO\n", __func__);
>   		goto release_out;
>   	}
> +
> +	WARN_ON_ONCE(bo->tbo.ttm->page_flags & TTM_PAGE_FLAG_SWAPPED);
> +
> +	memcpy(bo->tbo.ttm->pages,
> +			pages,
> +			sizeof(struct page*) * bo->tbo.ttm->num_pages);
>   	amdgpu_bo_placement_from_domain(bo, mem->domain);
>   	ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
>   	if (ret)
> @@ -872,6 +885,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>   release_out:
>   	amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
>   unregister_out:
> +	kvfree(pages);
>   	if (ret)
>   		amdgpu_mn_unregister(bo);
>   out:
Pan, Xinhui May 19, 2021, 4:09 a.m. UTC | #2
[AMD Official Use Only]

yes, we really dont swapout SG BOs.
The problems is that before we validate a userptr BO, we create this BO in CPU domain by default. So this BO has chance to swapout.

we set flag TTM_PAGE_FLAG_SG on userptr BO in popluate() which is too late.
I have not try to revert Chris' patch as I think it desnt help. Or I can have a try later.
Pan, Xinhui May 19, 2021, 5:07 a.m. UTC | #3
[AMD Official Use Only]

I have reverted Chris'  patch, still hit this failure.
Just see two lines in Chris' patch. Any BO in cpu domian would be swapout first. That is why we hit this issue frequently now. But the bug is there long time ago.

-       for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
-               list_for_each_entry(bo, &glob->swap_lru[i], swap) {
[snip]
+       for (i = TTM_PL_SYSTEM; i < TTM_NUM_MEM_TYPES; ++i) {
+               for (j = 0; j < TTM_MAX_BO_PRIORITY; ++j) {
Christian König May 19, 2021, 10:01 a.m. UTC | #4
I'm scratching my head how that is even possible.

See when a BO is created in the system domain it is just an empty hull, 
e.g. without backing store and allocated pages.

So the swapout function will just ignore it.

Christian.

Am 19.05.21 um 07:07 schrieb Pan, Xinhui:
> [AMD Official Use Only]
>
> I have reverted Chris'  patch, still hit this failure.
> Just see two lines in Chris' patch. Any BO in cpu domian would be swapout first. That is why we hit this issue frequently now. But the bug is there long time ago.
>
> -       for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
> -               list_for_each_entry(bo, &glob->swap_lru[i], swap) {
> [snip]
> +       for (i = TTM_PL_SYSTEM; i < TTM_NUM_MEM_TYPES; ++i) {
> +               for (j = 0; j < TTM_MAX_BO_PRIORITY; ++j) {
>
>
> ________________________________________
> 发件人: Pan, Xinhui <Xinhui.Pan@amd.com>
> 发送时间: 2021年5月19日 12:09
> 收件人: Kuehling, Felix; amd-gfx@lists.freedesktop.org
> 抄送: Deucher, Alexander; Koenig, Christian; dri-devel@lists.freedesktop.org; daniel@ffwll.ch
> 主题: 回复: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to swapout and swapin
>
> yes, we really dont swapout SG BOs.
> The problems is that before we validate a userptr BO, we create this BO in CPU domain by default. So this BO has chance to swapout.
>
> we set flag TTM_PAGE_FLAG_SG on userptr BO in popluate() which is too late.
> I have not try to revert Chris' patch as I think it desnt help. Or I can have a try later.
>
> ________________________________________
> 发件人: Kuehling, Felix <Felix.Kuehling@amd.com>
> 发送时间: 2021年5月19日 11:29
> 收件人: Pan, Xinhui; amd-gfx@lists.freedesktop.org
> 抄送: Deucher, Alexander; Koenig, Christian; dri-devel@lists.freedesktop.org; daniel@ffwll.ch
> 主题: Re: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to swapout and swapin
>
> Swapping SG BOs makes no sense, because TTM doesn't own the pages of
> this type of BO.
>
> Last I checked, userptr BOs (and other SG BOs) were protected from
> swapout by the fact that they would not be added to the swap-LRU. But it
> looks like Christian just removed the swap-LRU. I guess this broke that
> protection:
>
> commit 2cb51d22d70b18eaf339abf9758bf0b7608da65c
> Author: Christian König <christian.koenig@amd.com>
> Date:   Tue Oct 6 16:30:09 2020 +0200
>
>       drm/ttm: remove swap LRU v3
>
>       Instead evict round robin from each devices SYSTEM and TT domain.
>
>       v2: reorder num_pages access reported by Dan's script
>       v3: fix rebase fallout, num_pages should be 32bit
>
>       Signed-off-by: Christian König <christian.koenig@amd.com>
>       Tested-by: Nirmoy Das <nirmoy.das@amd.com>
>       Reviewed-by: Huang Rui <ray.huang@amd.com>
>       Reviewed-by: Matthew Auld <matthew.auld@intel.com>
>       Link: https://patchwork.freedesktop.org/patch/424009/
>
> Regards,
>     Felix
>
>
> On 2021-05-18 10:28 p.m., xinhui pan wrote:
>> cpu 1                                           cpu 2
>> kfd alloc BO A(userptr)                         alloc BO B(GTT)
>>       ->init -> validate                               -> init -> validate -> populate
>>       init_user_pages                            -> swapout BO A //hit ttm pages limit
>>        -> get_user_pages (fill up ttm->pages)
>>         -> validate -> populate
>>             -> swapin BO A // Now hit the BUG
>>
>> We know that get_user_pages may race with swapout on same BO.
>> Threre are some issues I have met.
>> 1) memory corruption.
>> This is because we do a swap before memory is setup. ttm_tt_swapout()
>> just create a swap_storage with its content being 0x0. So when we setup
>> memory after the swapout. The following swapin makes the memory
>> corrupted.
>>
>> 2) panic
>> When swapout happes with get_user_pages, they touch ttm->pages without
>> anylock. It causes memory corruption too. But I hit page fault mostly.
>>
>> Signed-off-by: xinhui pan <xinhui.pan@amd.com>
>> ---
>>    drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 16 +++++++++++++++-
>>    1 file changed, 15 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>> index 928e8d57cd08..42460e4480f8 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>> @@ -835,6 +835,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>>        struct amdkfd_process_info *process_info = mem->process_info;
>>        struct amdgpu_bo *bo = mem->bo;
>>        struct ttm_operation_ctx ctx = { true, false };
>> +     struct page **pages;
>>        int ret = 0;
>>
>>        mutex_lock(&process_info->lock);
>> @@ -852,7 +853,13 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>>                goto out;
>>        }
>>
>> -     ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages);
>> +     pages = kvmalloc_array(bo->tbo.ttm->num_pages,
>> +                     sizeof(struct page *),
>> +                     GFP_KERNEL | __GFP_ZERO);
>> +     if (!pages)
>> +             goto unregister_out;
>> +
>> +     ret = amdgpu_ttm_tt_get_user_pages(bo, pages);
>>        if (ret) {
>>                pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
>>                goto unregister_out;
>> @@ -863,6 +870,12 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>>                pr_err("%s: Failed to reserve BO\n", __func__);
>>                goto release_out;
>>        }
>> +
>> +     WARN_ON_ONCE(bo->tbo.ttm->page_flags & TTM_PAGE_FLAG_SWAPPED);
>> +
>> +     memcpy(bo->tbo.ttm->pages,
>> +                     pages,
>> +                     sizeof(struct page*) * bo->tbo.ttm->num_pages);
>>        amdgpu_bo_placement_from_domain(bo, mem->domain);
>>        ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
>>        if (ret)
>> @@ -872,6 +885,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>>    release_out:
>>        amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
>>    unregister_out:
>> +     kvfree(pages);
>>        if (ret)
>>                amdgpu_mn_unregister(bo);
>>    out:
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Felix Kuehling May 19, 2021, 3:11 p.m. UTC | #5
Looks like we're creating the userptr BO as ttm_bo_type_device. I guess
we should be using ttm_bo_type_sg? BTW, amdgpu_gem_userptr_ioctl also
uses ttm_bo_type_device.

Regards,
  Felix


Am 2021-05-19 um 6:01 a.m. schrieb Christian König:
> I'm scratching my head how that is even possible.
>
> See when a BO is created in the system domain it is just an empty
> hull, e.g. without backing store and allocated pages.
>
> So the swapout function will just ignore it.
>
> Christian.
>
> Am 19.05.21 um 07:07 schrieb Pan, Xinhui:
>> [AMD Official Use Only]
>>
>> I have reverted Chris'  patch, still hit this failure.
>> Just see two lines in Chris' patch. Any BO in cpu domian would be
>> swapout first. That is why we hit this issue frequently now. But the
>> bug is there long time ago.
>>
>> -       for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
>> -               list_for_each_entry(bo, &glob->swap_lru[i], swap) {
>> [snip]
>> +       for (i = TTM_PL_SYSTEM; i < TTM_NUM_MEM_TYPES; ++i) {
>> +               for (j = 0; j < TTM_MAX_BO_PRIORITY; ++j) {
>>
>>
>> ________________________________________
>> 发件人: Pan, Xinhui <Xinhui.Pan@amd.com>
>> 发送时间: 2021年5月19日 12:09
>> 收件人: Kuehling, Felix; amd-gfx@lists.freedesktop.org
>> 抄送: Deucher, Alexander; Koenig, Christian;
>> dri-devel@lists.freedesktop.org; daniel@ffwll.ch
>> 主题: 回复: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to
>> swapout and swapin
>>
>> yes, we really dont swapout SG BOs.
>> The problems is that before we validate a userptr BO, we create this
>> BO in CPU domain by default. So this BO has chance to swapout.
>>
>> we set flag TTM_PAGE_FLAG_SG on userptr BO in popluate() which is too
>> late.
>> I have not try to revert Chris' patch as I think it desnt help. Or I
>> can have a try later.
>>
>> ________________________________________
>> 发件人: Kuehling, Felix <Felix.Kuehling@amd.com>
>> 发送时间: 2021年5月19日 11:29
>> 收件人: Pan, Xinhui; amd-gfx@lists.freedesktop.org
>> 抄送: Deucher, Alexander; Koenig, Christian;
>> dri-devel@lists.freedesktop.org; daniel@ffwll.ch
>> 主题: Re: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to
>> swapout and swapin
>>
>> Swapping SG BOs makes no sense, because TTM doesn't own the pages of
>> this type of BO.
>>
>> Last I checked, userptr BOs (and other SG BOs) were protected from
>> swapout by the fact that they would not be added to the swap-LRU. But it
>> looks like Christian just removed the swap-LRU. I guess this broke that
>> protection:
>>
>> commit 2cb51d22d70b18eaf339abf9758bf0b7608da65c
>> Author: Christian König <christian.koenig@amd.com>
>> Date:   Tue Oct 6 16:30:09 2020 +0200
>>
>>       drm/ttm: remove swap LRU v3
>>
>>       Instead evict round robin from each devices SYSTEM and TT domain.
>>
>>       v2: reorder num_pages access reported by Dan's script
>>       v3: fix rebase fallout, num_pages should be 32bit
>>
>>       Signed-off-by: Christian König <christian.koenig@amd.com>
>>       Tested-by: Nirmoy Das <nirmoy.das@amd.com>
>>       Reviewed-by: Huang Rui <ray.huang@amd.com>
>>       Reviewed-by: Matthew Auld <matthew.auld@intel.com>
>>       Link: https://patchwork.freedesktop.org/patch/424009/
>>
>> Regards,
>>     Felix
>>
>>
>> On 2021-05-18 10:28 p.m., xinhui pan wrote:
>>> cpu 1                                           cpu 2
>>> kfd alloc BO A(userptr)                         alloc BO B(GTT)
>>>       ->init -> validate                               -> init ->
>>> validate -> populate
>>>       init_user_pages                            -> swapout BO A
>>> //hit ttm pages limit
>>>        -> get_user_pages (fill up ttm->pages)
>>>         -> validate -> populate
>>>             -> swapin BO A // Now hit the BUG
>>>
>>> We know that get_user_pages may race with swapout on same BO.
>>> Threre are some issues I have met.
>>> 1) memory corruption.
>>> This is because we do a swap before memory is setup. ttm_tt_swapout()
>>> just create a swap_storage with its content being 0x0. So when we setup
>>> memory after the swapout. The following swapin makes the memory
>>> corrupted.
>>>
>>> 2) panic
>>> When swapout happes with get_user_pages, they touch ttm->pages without
>>> anylock. It causes memory corruption too. But I hit page fault mostly.
>>>
>>> Signed-off-by: xinhui pan <xinhui.pan@amd.com>
>>> ---
>>>    drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 16
>>> +++++++++++++++-
>>>    1 file changed, 15 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>> index 928e8d57cd08..42460e4480f8 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>> @@ -835,6 +835,7 @@ static int init_user_pages(struct kgd_mem *mem,
>>> uint64_t user_addr)
>>>        struct amdkfd_process_info *process_info = mem->process_info;
>>>        struct amdgpu_bo *bo = mem->bo;
>>>        struct ttm_operation_ctx ctx = { true, false };
>>> +     struct page **pages;
>>>        int ret = 0;
>>>
>>>        mutex_lock(&process_info->lock);
>>> @@ -852,7 +853,13 @@ static int init_user_pages(struct kgd_mem *mem,
>>> uint64_t user_addr)
>>>                goto out;
>>>        }
>>>
>>> -     ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages);
>>> +     pages = kvmalloc_array(bo->tbo.ttm->num_pages,
>>> +                     sizeof(struct page *),
>>> +                     GFP_KERNEL | __GFP_ZERO);
>>> +     if (!pages)
>>> +             goto unregister_out;
>>> +
>>> +     ret = amdgpu_ttm_tt_get_user_pages(bo, pages);
>>>        if (ret) {
>>>                pr_err("%s: Failed to get user pages: %d\n",
>>> __func__, ret);
>>>                goto unregister_out;
>>> @@ -863,6 +870,12 @@ static int init_user_pages(struct kgd_mem *mem,
>>> uint64_t user_addr)
>>>                pr_err("%s: Failed to reserve BO\n", __func__);
>>>                goto release_out;
>>>        }
>>> +
>>> +     WARN_ON_ONCE(bo->tbo.ttm->page_flags & TTM_PAGE_FLAG_SWAPPED);
>>> +
>>> +     memcpy(bo->tbo.ttm->pages,
>>> +                     pages,
>>> +                     sizeof(struct page*) * bo->tbo.ttm->num_pages);
>>>        amdgpu_bo_placement_from_domain(bo, mem->domain);
>>>        ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
>>>        if (ret)
>>> @@ -872,6 +885,7 @@ static int init_user_pages(struct kgd_mem *mem,
>>> uint64_t user_addr)
>>>    release_out:
>>>        amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
>>>    unregister_out:
>>> +     kvfree(pages);
>>>        if (ret)
>>>                amdgpu_mn_unregister(bo);
>>>    out:
>> _______________________________________________
>> amd-gfx mailing list
>> amd-gfx@lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
>
Pan, Xinhui May 20, 2021, 2:55 a.m. UTC | #6
[AMD Official Use Only]

swapout function create one swap storage which is filled with zero. And set ttm->page_flags as TTM_PAGE_FLAG_SWAPPED.  Just because ttm has no backend page this time, no real data is swapout to this swap storage.

swapin function is called during populate as TTM_PAGE_FLAG_SWAPPED is set.
Now here is the problem, we swapin data to ttm bakend memory from swap storage. That just causes the memory been overwritten.
Pan, Xinhui May 20, 2021, 2:58 a.m. UTC | #7
[AMD Official Use Only]

I am not sure if we can create a ttm_bo_type_sg bo for userptr. But I have another idea now. we can use flag AMDGPU_AMDKFD_CREATE_USERPTR_BO to create the userptr bo.
Christian König May 20, 2021, 6:38 a.m. UTC | #8
The problem is that ttm_bo_type_sg doesn't allocate a page array for the 
TT object.

Christian.

Am 20.05.21 um 04:58 schrieb Pan, Xinhui:
> [AMD Official Use Only]
>
> I am not sure if we can create a ttm_bo_type_sg bo for userptr. But I have another idea now. we can use flag AMDGPU_AMDKFD_CREATE_USERPTR_BO to create the userptr bo.
> ________________________________________
> 发件人: Kuehling, Felix <Felix.Kuehling@amd.com>
> 发送时间: 2021年5月19日 23:11
> 收件人: Christian König; Pan, Xinhui; amd-gfx@lists.freedesktop.org
> 抄送: Deucher, Alexander; daniel@ffwll.ch; Koenig, Christian; dri-devel@lists.freedesktop.org
> 主题: Re: 回复: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to swapout and swapin
>
> Looks like we're creating the userptr BO as ttm_bo_type_device. I guess
> we should be using ttm_bo_type_sg? BTW, amdgpu_gem_userptr_ioctl also
> uses ttm_bo_type_device.
>
> Regards,
>    Felix
>
>
> Am 2021-05-19 um 6:01 a.m. schrieb Christian König:
>> I'm scratching my head how that is even possible.
>>
>> See when a BO is created in the system domain it is just an empty
>> hull, e.g. without backing store and allocated pages.
>>
>> So the swapout function will just ignore it.
>>
>> Christian.
>>
>> Am 19.05.21 um 07:07 schrieb Pan, Xinhui:
>>> [AMD Official Use Only]
>>>
>>> I have reverted Chris'  patch, still hit this failure.
>>> Just see two lines in Chris' patch. Any BO in cpu domian would be
>>> swapout first. That is why we hit this issue frequently now. But the
>>> bug is there long time ago.
>>>
>>> -       for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
>>> -               list_for_each_entry(bo, &glob->swap_lru[i], swap) {
>>> [snip]
>>> +       for (i = TTM_PL_SYSTEM; i < TTM_NUM_MEM_TYPES; ++i) {
>>> +               for (j = 0; j < TTM_MAX_BO_PRIORITY; ++j) {
>>>
>>>
>>> ________________________________________
>>> 发件人: Pan, Xinhui <Xinhui.Pan@amd.com>
>>> 发送时间: 2021年5月19日 12:09
>>> 收件人: Kuehling, Felix; amd-gfx@lists.freedesktop.org
>>> 抄送: Deucher, Alexander; Koenig, Christian;
>>> dri-devel@lists.freedesktop.org; daniel@ffwll.ch
>>> 主题: 回复: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to
>>> swapout and swapin
>>>
>>> yes, we really dont swapout SG BOs.
>>> The problems is that before we validate a userptr BO, we create this
>>> BO in CPU domain by default. So this BO has chance to swapout.
>>>
>>> we set flag TTM_PAGE_FLAG_SG on userptr BO in popluate() which is too
>>> late.
>>> I have not try to revert Chris' patch as I think it desnt help. Or I
>>> can have a try later.
>>>
>>> ________________________________________
>>> 发件人: Kuehling, Felix <Felix.Kuehling@amd.com>
>>> 发送时间: 2021年5月19日 11:29
>>> 收件人: Pan, Xinhui; amd-gfx@lists.freedesktop.org
>>> 抄送: Deucher, Alexander; Koenig, Christian;
>>> dri-devel@lists.freedesktop.org; daniel@ffwll.ch
>>> 主题: Re: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to
>>> swapout and swapin
>>>
>>> Swapping SG BOs makes no sense, because TTM doesn't own the pages of
>>> this type of BO.
>>>
>>> Last I checked, userptr BOs (and other SG BOs) were protected from
>>> swapout by the fact that they would not be added to the swap-LRU. But it
>>> looks like Christian just removed the swap-LRU. I guess this broke that
>>> protection:
>>>
>>> commit 2cb51d22d70b18eaf339abf9758bf0b7608da65c
>>> Author: Christian König <christian.koenig@amd.com>
>>> Date:   Tue Oct 6 16:30:09 2020 +0200
>>>
>>>        drm/ttm: remove swap LRU v3
>>>
>>>        Instead evict round robin from each devices SYSTEM and TT domain.
>>>
>>>        v2: reorder num_pages access reported by Dan's script
>>>        v3: fix rebase fallout, num_pages should be 32bit
>>>
>>>        Signed-off-by: Christian König <christian.koenig@amd.com>
>>>        Tested-by: Nirmoy Das <nirmoy.das@amd.com>
>>>        Reviewed-by: Huang Rui <ray.huang@amd.com>
>>>        Reviewed-by: Matthew Auld <matthew.auld@intel.com>
>>>        Link: https://patchwork.freedesktop.org/patch/424009/
>>>
>>> Regards,
>>>      Felix
>>>
>>>
>>> On 2021-05-18 10:28 p.m., xinhui pan wrote:
>>>> cpu 1                                           cpu 2
>>>> kfd alloc BO A(userptr)                         alloc BO B(GTT)
>>>>        ->init -> validate                               -> init ->
>>>> validate -> populate
>>>>        init_user_pages                            -> swapout BO A
>>>> //hit ttm pages limit
>>>>         -> get_user_pages (fill up ttm->pages)
>>>>          -> validate -> populate
>>>>              -> swapin BO A // Now hit the BUG
>>>>
>>>> We know that get_user_pages may race with swapout on same BO.
>>>> Threre are some issues I have met.
>>>> 1) memory corruption.
>>>> This is because we do a swap before memory is setup. ttm_tt_swapout()
>>>> just create a swap_storage with its content being 0x0. So when we setup
>>>> memory after the swapout. The following swapin makes the memory
>>>> corrupted.
>>>>
>>>> 2) panic
>>>> When swapout happes with get_user_pages, they touch ttm->pages without
>>>> anylock. It causes memory corruption too. But I hit page fault mostly.
>>>>
>>>> Signed-off-by: xinhui pan <xinhui.pan@amd.com>
>>>> ---
>>>>     drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 16
>>>> +++++++++++++++-
>>>>     1 file changed, 15 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>>> index 928e8d57cd08..42460e4480f8 100644
>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>>> @@ -835,6 +835,7 @@ static int init_user_pages(struct kgd_mem *mem,
>>>> uint64_t user_addr)
>>>>         struct amdkfd_process_info *process_info = mem->process_info;
>>>>         struct amdgpu_bo *bo = mem->bo;
>>>>         struct ttm_operation_ctx ctx = { true, false };
>>>> +     struct page **pages;
>>>>         int ret = 0;
>>>>
>>>>         mutex_lock(&process_info->lock);
>>>> @@ -852,7 +853,13 @@ static int init_user_pages(struct kgd_mem *mem,
>>>> uint64_t user_addr)
>>>>                 goto out;
>>>>         }
>>>>
>>>> -     ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages);
>>>> +     pages = kvmalloc_array(bo->tbo.ttm->num_pages,
>>>> +                     sizeof(struct page *),
>>>> +                     GFP_KERNEL | __GFP_ZERO);
>>>> +     if (!pages)
>>>> +             goto unregister_out;
>>>> +
>>>> +     ret = amdgpu_ttm_tt_get_user_pages(bo, pages);
>>>>         if (ret) {
>>>>                 pr_err("%s: Failed to get user pages: %d\n",
>>>> __func__, ret);
>>>>                 goto unregister_out;
>>>> @@ -863,6 +870,12 @@ static int init_user_pages(struct kgd_mem *mem,
>>>> uint64_t user_addr)
>>>>                 pr_err("%s: Failed to reserve BO\n", __func__);
>>>>                 goto release_out;
>>>>         }
>>>> +
>>>> +     WARN_ON_ONCE(bo->tbo.ttm->page_flags & TTM_PAGE_FLAG_SWAPPED);
>>>> +
>>>> +     memcpy(bo->tbo.ttm->pages,
>>>> +                     pages,
>>>> +                     sizeof(struct page*) * bo->tbo.ttm->num_pages);
>>>>         amdgpu_bo_placement_from_domain(bo, mem->domain);
>>>>         ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
>>>>         if (ret)
>>>> @@ -872,6 +885,7 @@ static int init_user_pages(struct kgd_mem *mem,
>>>> uint64_t user_addr)
>>>>     release_out:
>>>>         amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
>>>>     unregister_out:
>>>> +     kvfree(pages);
>>>>         if (ret)
>>>>                 amdgpu_mn_unregister(bo);
>>>>     out:
>>> _______________________________________________
>>> amd-gfx mailing list
>>> amd-gfx@lists.freedesktop.org
>>> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Christian König May 20, 2021, 6:39 a.m. UTC | #9
> swapout function create one swap storage which is filled with zero. And set ttm->page_flags as TTM_PAGE_FLAG_SWAPPED.  Just because ttm has no backend page this time, no real data is swapout to this swap storage.

That's the fundamental problem. A TT object which isn't populated 
shouldn't be considered for swapout nor eviction in the first place.

I'm going to take a look later today.

Christian.

Am 20.05.21 um 04:55 schrieb Pan, Xinhui:
> [AMD Official Use Only]
>
> swapout function create one swap storage which is filled with zero. And set ttm->page_flags as TTM_PAGE_FLAG_SWAPPED.  Just because ttm has no backend page this time, no real data is swapout to this swap storage.
>
> swapin function is called during populate as TTM_PAGE_FLAG_SWAPPED is set.
> Now here is the problem, we swapin data to ttm bakend memory from swap storage. That just causes the memory been overwritten.
>
> ________________________________________
> 发件人: Christian König <ckoenig.leichtzumerken@gmail.com>
> 发送时间: 2021年5月19日 18:01
> 收件人: Pan, Xinhui; Kuehling, Felix; amd-gfx@lists.freedesktop.org
> 抄送: Deucher, Alexander; daniel@ffwll.ch; Koenig, Christian; dri-devel@lists.freedesktop.org
> 主题: Re: 回复: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to swapout and swapin
>
> I'm scratching my head how that is even possible.
>
> See when a BO is created in the system domain it is just an empty hull,
> e.g. without backing store and allocated pages.
>
> So the swapout function will just ignore it.
>
> Christian.
>
> Am 19.05.21 um 07:07 schrieb Pan, Xinhui:
>> [AMD Official Use Only]
>>
>> I have reverted Chris'  patch, still hit this failure.
>> Just see two lines in Chris' patch. Any BO in cpu domian would be swapout first. That is why we hit this issue frequently now. But the bug is there long time ago.
>>
>> -       for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
>> -               list_for_each_entry(bo, &glob->swap_lru[i], swap) {
>> [snip]
>> +       for (i = TTM_PL_SYSTEM; i < TTM_NUM_MEM_TYPES; ++i) {
>> +               for (j = 0; j < TTM_MAX_BO_PRIORITY; ++j) {
>>
>>
>> ________________________________________
>> 发件人: Pan, Xinhui <Xinhui.Pan@amd.com>
>> 发送时间: 2021年5月19日 12:09
>> 收件人: Kuehling, Felix; amd-gfx@lists.freedesktop.org
>> 抄送: Deucher, Alexander; Koenig, Christian; dri-devel@lists.freedesktop.org; daniel@ffwll.ch
>> 主题: 回复: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to swapout and swapin
>>
>> yes, we really dont swapout SG BOs.
>> The problems is that before we validate a userptr BO, we create this BO in CPU domain by default. So this BO has chance to swapout.
>>
>> we set flag TTM_PAGE_FLAG_SG on userptr BO in popluate() which is too late.
>> I have not try to revert Chris' patch as I think it desnt help. Or I can have a try later.
>>
>> ________________________________________
>> 发件人: Kuehling, Felix <Felix.Kuehling@amd.com>
>> 发送时间: 2021年5月19日 11:29
>> 收件人: Pan, Xinhui; amd-gfx@lists.freedesktop.org
>> 抄送: Deucher, Alexander; Koenig, Christian; dri-devel@lists.freedesktop.org; daniel@ffwll.ch
>> 主题: Re: [RFC PATCH 1/2] drm/amdgpu: Fix memory corruption due to swapout and swapin
>>
>> Swapping SG BOs makes no sense, because TTM doesn't own the pages of
>> this type of BO.
>>
>> Last I checked, userptr BOs (and other SG BOs) were protected from
>> swapout by the fact that they would not be added to the swap-LRU. But it
>> looks like Christian just removed the swap-LRU. I guess this broke that
>> protection:
>>
>> commit 2cb51d22d70b18eaf339abf9758bf0b7608da65c
>> Author: Christian König <christian.koenig@amd.com>
>> Date:   Tue Oct 6 16:30:09 2020 +0200
>>
>>        drm/ttm: remove swap LRU v3
>>
>>        Instead evict round robin from each devices SYSTEM and TT domain.
>>
>>        v2: reorder num_pages access reported by Dan's script
>>        v3: fix rebase fallout, num_pages should be 32bit
>>
>>        Signed-off-by: Christian König <christian.koenig@amd.com>
>>        Tested-by: Nirmoy Das <nirmoy.das@amd.com>
>>        Reviewed-by: Huang Rui <ray.huang@amd.com>
>>        Reviewed-by: Matthew Auld <matthew.auld@intel.com>
>>        Link: https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpatchwork.freedesktop.org%2Fpatch%2F424009%2F&amp;data=04%7C01%7CXinhui.Pan%40amd.com%7Cb4422d8b3e4947cd243c08d91aad14c3%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637570152942496679%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=K3%2FnFpN56y8L49UuYRM6SqefVFLnqIwpDAtWpS1XvnQ%3D&amp;reserved=0
>>
>> Regards,
>>      Felix
>>
>>
>> On 2021-05-18 10:28 p.m., xinhui pan wrote:
>>> cpu 1                                           cpu 2
>>> kfd alloc BO A(userptr)                         alloc BO B(GTT)
>>>        ->init -> validate                               -> init -> validate -> populate
>>>        init_user_pages                            -> swapout BO A //hit ttm pages limit
>>>         -> get_user_pages (fill up ttm->pages)
>>>          -> validate -> populate
>>>              -> swapin BO A // Now hit the BUG
>>>
>>> We know that get_user_pages may race with swapout on same BO.
>>> Threre are some issues I have met.
>>> 1) memory corruption.
>>> This is because we do a swap before memory is setup. ttm_tt_swapout()
>>> just create a swap_storage with its content being 0x0. So when we setup
>>> memory after the swapout. The following swapin makes the memory
>>> corrupted.
>>>
>>> 2) panic
>>> When swapout happes with get_user_pages, they touch ttm->pages without
>>> anylock. It causes memory corruption too. But I hit page fault mostly.
>>>
>>> Signed-off-by: xinhui pan <xinhui.pan@amd.com>
>>> ---
>>>     drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 16 +++++++++++++++-
>>>     1 file changed, 15 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>> index 928e8d57cd08..42460e4480f8 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>>> @@ -835,6 +835,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>>>         struct amdkfd_process_info *process_info = mem->process_info;
>>>         struct amdgpu_bo *bo = mem->bo;
>>>         struct ttm_operation_ctx ctx = { true, false };
>>> +     struct page **pages;
>>>         int ret = 0;
>>>
>>>         mutex_lock(&process_info->lock);
>>> @@ -852,7 +853,13 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>>>                 goto out;
>>>         }
>>>
>>> -     ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages);
>>> +     pages = kvmalloc_array(bo->tbo.ttm->num_pages,
>>> +                     sizeof(struct page *),
>>> +                     GFP_KERNEL | __GFP_ZERO);
>>> +     if (!pages)
>>> +             goto unregister_out;
>>> +
>>> +     ret = amdgpu_ttm_tt_get_user_pages(bo, pages);
>>>         if (ret) {
>>>                 pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
>>>                 goto unregister_out;
>>> @@ -863,6 +870,12 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>>>                 pr_err("%s: Failed to reserve BO\n", __func__);
>>>                 goto release_out;
>>>         }
>>> +
>>> +     WARN_ON_ONCE(bo->tbo.ttm->page_flags & TTM_PAGE_FLAG_SWAPPED);
>>> +
>>> +     memcpy(bo->tbo.ttm->pages,
>>> +                     pages,
>>> +                     sizeof(struct page*) * bo->tbo.ttm->num_pages);
>>>         amdgpu_bo_placement_from_domain(bo, mem->domain);
>>>         ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
>>>         if (ret)
>>> @@ -872,6 +885,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
>>>     release_out:
>>>         amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
>>>     unregister_out:
>>> +     kvfree(pages);
>>>         if (ret)
>>>                 amdgpu_mn_unregister(bo);
>>>     out:
>> _______________________________________________
>> amd-gfx mailing list
>> amd-gfx@lists.freedesktop.org
>> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=04%7C01%7CXinhui.Pan%40amd.com%7Cb4422d8b3e4947cd243c08d91aad14c3%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637570152942496679%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=QOuCH8r6DMdp%2Bvr6LAPhDw0BUCKPzK%2Bbv6MoHx4AmMo%3D&amp;reserved=0
Pan, Xinhui May 20, 2021, 7:30 a.m. UTC | #10
I just sent out patch below yesterday.  swapping unpopulated bo is useless indeed.

[RFC PATCH 2/2] drm/ttm: skip swapout when ttm has no backend page.
Pan, Xinhui May 20, 2021, 7:32 a.m. UTC | #11
[AMD Official Use Only]

I just sent out patch below yesterday.  swapping unpopulated bo is useless indeed.

[RFC PATCH 2/2] drm/ttm: skip swapout when ttm has no backend page.
diff mbox series

Patch

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 928e8d57cd08..42460e4480f8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -835,6 +835,7 @@  static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
 	struct amdkfd_process_info *process_info = mem->process_info;
 	struct amdgpu_bo *bo = mem->bo;
 	struct ttm_operation_ctx ctx = { true, false };
+	struct page **pages;
 	int ret = 0;
 
 	mutex_lock(&process_info->lock);
@@ -852,7 +853,13 @@  static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
 		goto out;
 	}
 
-	ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages);
+	pages = kvmalloc_array(bo->tbo.ttm->num_pages,
+			sizeof(struct page *),
+			GFP_KERNEL | __GFP_ZERO);
+	if (!pages)
+		goto unregister_out;
+
+	ret = amdgpu_ttm_tt_get_user_pages(bo, pages);
 	if (ret) {
 		pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
 		goto unregister_out;
@@ -863,6 +870,12 @@  static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
 		pr_err("%s: Failed to reserve BO\n", __func__);
 		goto release_out;
 	}
+
+	WARN_ON_ONCE(bo->tbo.ttm->page_flags & TTM_PAGE_FLAG_SWAPPED);
+
+	memcpy(bo->tbo.ttm->pages,
+			pages,
+			sizeof(struct page*) * bo->tbo.ttm->num_pages);
 	amdgpu_bo_placement_from_domain(bo, mem->domain);
 	ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 	if (ret)
@@ -872,6 +885,7 @@  static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr)
 release_out:
 	amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
 unregister_out:
+	kvfree(pages);
 	if (ret)
 		amdgpu_mn_unregister(bo);
 out: