From: Nirmoy Das <nirmoy.das@linux.intel.com>
To: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: "Matthew Auld" <matthew.auld@intel.com>,
"Nirmoy Das" <nirmoy.das@intel.com>,
intel-xe@lists.freedesktop.org,
"Thomas Hellström" <thomas.hellstrom@linux.intel.com>,
"Matthew Brost" <matthew.brost@intel.com>,
stable@vger.kernel.org
Subject: Re: [PATCH v2 2/2] drm/xe: Wait for migration job before unmapping pages
Date: Tue, 10 Dec 2024 16:59:28 +0100 [thread overview]
Message-ID: <e9d7a359-2c38-4dab-a56e-b406198cf033@linux.intel.com> (raw)
In-Reply-To: <6vthulgckq2jpfwn2amux5ssijwhxbjp44g2xwp44r626ttrkh@ofwboiylc7eh>
On 12/10/2024 4:38 PM, Lucas De Marchi wrote:
> On Thu, Dec 05, 2024 at 03:33:46PM +0100, Nirmoy Das wrote:
>>
>> On 12/5/2024 1:40 PM, Matthew Auld wrote:
>>> On 05/12/2024 12:02, Nirmoy Das wrote:
>>>> There could be still migration job going on while doing
>>>> xe_tt_unmap_sg() which could trigger GPU page faults. Fix this by
>>>> waiting for the migration job to finish.
>>>>
>>>> v2: Use intr=false(Matt A)
>>>>
>>>> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/3466
>>>> Fixes: 75521e8b56e8 ("drm/xe: Perform dma_map when moving system buffer objects to TT")
>>>> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
>>>> Cc: Matthew Brost <matthew.brost@intel.com>
>>>> Cc: Lucas De Marchi <lucas.demarchi@intel.com>
>>>> Cc: <stable@vger.kernel.org> # v6.11+
>>>> Cc: Matthew Auld <matthew.auld@intel.com>
>>>> Signed-off-by: Nirmoy Das <nirmoy.das@intel.com>
>>>
>>> Ok, so this is something like ttm_bo_move_to_ghost() doing a pipeline move for tt -> system, but we then do xe_tt_unmap_sg() too early which tears down the IOMMU (if enabled) mappings whilst the job is in progress?
>> Yes, this exactly what is happening for this issue.
>>>
>>> Maybe add some more info to the commit message?
>>
>> I will add more details.
>
> Are you going to send a new version?
Was waiting for more reviews. Sent out v3 with updated commit message.
> Once this is fixed, please also
> send a revert MR to the kconfig workaround
> 3940181b1bad @ gitlab.freedesktop.org/drm/xe/ci.git
I will do that.
Thanks,
Nirmoy
>
> Lucas De Marchi
>
>>
>>
>>> I think this for sure fixes it. Just wondering if it's somehow possible to keep the mapping until the job is done, since all tt -> sys moves are now synced here?
>>>
>>> Unless Thomas has a better idea here,
>>> Reviewed-by: Matthew Auld <matthew.auld@intel.com>
>>
>>
>> Thanks,
>>
>> Nirmoy
>>
>>>
>>>> ---
>>>> drivers/gpu/drm/xe/xe_bo.c | 10 +++++++++-
>>>> 1 file changed, 9 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
>>>> index b2aa368a23f8..c906a5529db0 100644
>>>> --- a/drivers/gpu/drm/xe/xe_bo.c
>>>> +++ b/drivers/gpu/drm/xe/xe_bo.c
>>>> @@ -857,8 +857,16 @@ static int xe_bo_move(struct ttm_buffer_object *ttm_bo, bool evict,
>>>> out:
>>>> if ((!ttm_bo->resource || ttm_bo->resource->mem_type == XE_PL_SYSTEM) &&
>>>> - ttm_bo->ttm)
>>>> + ttm_bo->ttm) {
>>>> + long timeout = dma_resv_wait_timeout(ttm_bo->base.resv,
>>>> + DMA_RESV_USAGE_BOOKKEEP,
>>>> + false,
>>>> + MAX_SCHEDULE_TIMEOUT);
>>>> + if (timeout < 0)
>>>> + ret = timeout;
>>>> +
>>>> xe_tt_unmap_sg(ttm_bo->ttm);
>>>> + }
>>>> return ret;
>>>> }
>>>
next prev parent reply other threads:[~2024-12-10 15:59 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-12-05 12:02 [PATCH v2 1/2] drm/xe: Use non-interruptible wait when moving BO to system Nirmoy Das
2024-12-05 11:55 ` Matthew Auld
2024-12-05 12:02 ` [PATCH v2 2/2] drm/xe: Wait for migration job before unmapping pages Nirmoy Das
2024-12-05 12:40 ` Matthew Auld
2024-12-05 14:33 ` Nirmoy Das
2024-12-10 15:38 ` Lucas De Marchi
2024-12-10 15:59 ` Nirmoy Das [this message]
2024-12-10 16:52 ` Thomas Hellström
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=e9d7a359-2c38-4dab-a56e-b406198cf033@linux.intel.com \
--to=nirmoy.das@linux.intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=lucas.demarchi@intel.com \
--cc=matthew.auld@intel.com \
--cc=matthew.brost@intel.com \
--cc=nirmoy.das@intel.com \
--cc=stable@vger.kernel.org \
--cc=thomas.hellstrom@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox