AMD-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Christian König" <christian.koenig@amd.com>
To: Alex Deucher <alexdeucher@gmail.com>
Cc: Nikita Zhandarovich <n.zhandarovich@fintech.ru>,
	Alex Deucher <alexander.deucher@amd.com>,
	Xinhui Pan <Xinhui.Pan@amd.com>, David Airlie <airlied@gmail.com>,
	Daniel Vetter <daniel@ffwll.ch>,
	Jerome Glisse <jglisse@redhat.com>,
	Dave Airlie <airlied@redhat.com>,
	amd-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org,
	linux-kernel@vger.kernel.org, lvc-project@linuxtesting.org,
	stable@vger.kernel.org
Subject: Re: [PATCH] drm/radeon/evergreen_cs: fix int overflow errors in cs track offsets
Date: Mon, 29 Jul 2024 11:23:00 +0200	[thread overview]
Message-ID: <fb530f45-df88-402a-9dc0-99298b88754c@amd.com> (raw)
In-Reply-To: <CADnq5_PuzU12x=M09HaGkG7Yqg8Lk1M1nWDAut7iP09TT33D6g@mail.gmail.com>

Am 26.07.24 um 14:52 schrieb Alex Deucher:
> On Fri, Jul 26, 2024 at 3:05 AM Christian König
> <christian.koenig@amd.com> wrote:
>> Am 25.07.24 um 20:09 schrieb Nikita Zhandarovich:
>>> Several cs track offsets (such as 'track->db_s_read_offset')
>>> either are initialized with or plainly take big enough values that,
>>> once shifted 8 bits left, may be hit with integer overflow if the
>>> resulting values end up going over u32 limit.
>>>
>>> Some debug prints take this into account (see according dev_warn() in
>>> evergreen_cs_track_validate_stencil()), even if the actual
>>> calculated value assigned to local 'offset' variable is missing
>>> similar proper expansion.
>>>
>>> Mitigate the problem by casting the type of right operands to the
>>> wider type of corresponding left ones in all such cases.
>>>
>>> Found by Linux Verification Center (linuxtesting.org) with static
>>> analysis tool SVACE.
>>>
>>> Fixes: 285484e2d55e ("drm/radeon: add support for evergreen/ni tiling informations v11")
>>> Cc: stable@vger.kernel.org
>> Well first of all the long cast doesn't makes the value 64bit, it
>> depends on the architecture.
>>
>> Then IIRC the underlying hw can only handle a 32bit address space so
>> having the offset as long is incorrect to begin with.
> Evergreen chips support a 36 bit internal address space and NI and
> newer support a 40 bit one, so this is applicable.

In that case I strongly suggest that we replace the unsigned long with 
u64 or otherwise we get different behavior on 32 and 64bit machines.

Regards,
Christian.

>
> Alex
>
>> And finally that is absolutely not material for stable.
>>
>> Regards,
>> Christian.
>>
>>> Signed-off-by: Nikita Zhandarovich <n.zhandarovich@fintech.ru>
>>> ---
>>> P.S. While I am not certain that track->cb_color_bo_offset[id]
>>> actually ends up taking values high enough to cause an overflow,
>>> nonetheless I thought it prudent to cast it to ulong as well.
>>>
>>>    drivers/gpu/drm/radeon/evergreen_cs.c | 18 +++++++++---------
>>>    1 file changed, 9 insertions(+), 9 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/radeon/evergreen_cs.c b/drivers/gpu/drm/radeon/evergreen_cs.c
>>> index 1fe6e0d883c7..d734d221e2da 100644
>>> --- a/drivers/gpu/drm/radeon/evergreen_cs.c
>>> +++ b/drivers/gpu/drm/radeon/evergreen_cs.c
>>> @@ -433,7 +433,7 @@ static int evergreen_cs_track_validate_cb(struct radeon_cs_parser *p, unsigned i
>>>                return r;
>>>        }
>>>
>>> -     offset = track->cb_color_bo_offset[id] << 8;
>>> +     offset = (unsigned long)track->cb_color_bo_offset[id] << 8;
>>>        if (offset & (surf.base_align - 1)) {
>>>                dev_warn(p->dev, "%s:%d cb[%d] bo base %ld not aligned with %ld\n",
>>>                         __func__, __LINE__, id, offset, surf.base_align);
>>> @@ -455,7 +455,7 @@ static int evergreen_cs_track_validate_cb(struct radeon_cs_parser *p, unsigned i
>>>                                min = surf.nby - 8;
>>>                        }
>>>                        bsize = radeon_bo_size(track->cb_color_bo[id]);
>>> -                     tmp = track->cb_color_bo_offset[id] << 8;
>>> +                     tmp = (unsigned long)track->cb_color_bo_offset[id] << 8;
>>>                        for (nby = surf.nby; nby > min; nby--) {
>>>                                size = nby * surf.nbx * surf.bpe * surf.nsamples;
>>>                                if ((tmp + size * mslice) <= bsize) {
>>> @@ -476,10 +476,10 @@ static int evergreen_cs_track_validate_cb(struct radeon_cs_parser *p, unsigned i
>>>                        }
>>>                }
>>>                dev_warn(p->dev, "%s:%d cb[%d] bo too small (layer size %d, "
>>> -                      "offset %d, max layer %d, bo size %ld, slice %d)\n",
>>> +                      "offset %ld, max layer %d, bo size %ld, slice %d)\n",
>>>                         __func__, __LINE__, id, surf.layer_size,
>>> -                     track->cb_color_bo_offset[id] << 8, mslice,
>>> -                     radeon_bo_size(track->cb_color_bo[id]), slice);
>>> +                     (unsigned long)track->cb_color_bo_offset[id] << 8,
>>> +                     mslice, radeon_bo_size(track->cb_color_bo[id]), slice);
>>>                dev_warn(p->dev, "%s:%d problematic surf: (%d %d) (%d %d %d %d %d %d %d)\n",
>>>                         __func__, __LINE__, surf.nbx, surf.nby,
>>>                        surf.mode, surf.bpe, surf.nsamples,
>>> @@ -608,7 +608,7 @@ static int evergreen_cs_track_validate_stencil(struct radeon_cs_parser *p)
>>>                return r;
>>>        }
>>>
>>> -     offset = track->db_s_read_offset << 8;
>>> +     offset = (unsigned long)track->db_s_read_offset << 8;
>>>        if (offset & (surf.base_align - 1)) {
>>>                dev_warn(p->dev, "%s:%d stencil read bo base %ld not aligned with %ld\n",
>>>                         __func__, __LINE__, offset, surf.base_align);
>>> @@ -627,7 +627,7 @@ static int evergreen_cs_track_validate_stencil(struct radeon_cs_parser *p)
>>>                return -EINVAL;
>>>        }
>>>
>>> -     offset = track->db_s_write_offset << 8;
>>> +     offset = (unsigned long)track->db_s_write_offset << 8;
>>>        if (offset & (surf.base_align - 1)) {
>>>                dev_warn(p->dev, "%s:%d stencil write bo base %ld not aligned with %ld\n",
>>>                         __func__, __LINE__, offset, surf.base_align);
>>> @@ -706,7 +706,7 @@ static int evergreen_cs_track_validate_depth(struct radeon_cs_parser *p)
>>>                return r;
>>>        }
>>>
>>> -     offset = track->db_z_read_offset << 8;
>>> +     offset = (unsigned long)track->db_z_read_offset << 8;
>>>        if (offset & (surf.base_align - 1)) {
>>>                dev_warn(p->dev, "%s:%d stencil read bo base %ld not aligned with %ld\n",
>>>                         __func__, __LINE__, offset, surf.base_align);
>>> @@ -722,7 +722,7 @@ static int evergreen_cs_track_validate_depth(struct radeon_cs_parser *p)
>>>                return -EINVAL;
>>>        }
>>>
>>> -     offset = track->db_z_write_offset << 8;
>>> +     offset = (unsigned long)track->db_z_write_offset << 8;
>>>        if (offset & (surf.base_align - 1)) {
>>>                dev_warn(p->dev, "%s:%d stencil write bo base %ld not aligned with %ld\n",
>>>                         __func__, __LINE__, offset, surf.base_align);


  reply	other threads:[~2024-07-29  9:23 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-07-25 18:09 [PATCH] drm/radeon/evergreen_cs: fix int overflow errors in cs track offsets Nikita Zhandarovich
2024-07-25 20:59 ` Alex Deucher
2024-07-26  7:06   ` Christian König
2024-07-26  7:05 ` Christian König
2024-07-26 12:52   ` Alex Deucher
2024-07-29  9:23     ` Christian König [this message]
2024-07-29 17:26       ` Nikita Zhandarovich
2024-07-29 18:04         ` Christian König
2024-07-29 18:12           ` Christian König
2024-07-30 17:36             ` Nikita Zhandarovich
2024-07-31  6:56               ` Christian König
2024-08-05  7:34                 ` Nikita Zhandarovich
2024-08-05 14:36                   ` Alex Deucher
2024-08-19  8:23                   ` Christian König

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=fb530f45-df88-402a-9dc0-99298b88754c@amd.com \
    --to=christian.koenig@amd.com \
    --cc=Xinhui.Pan@amd.com \
    --cc=airlied@gmail.com \
    --cc=airlied@redhat.com \
    --cc=alexander.deucher@amd.com \
    --cc=alexdeucher@gmail.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=daniel@ffwll.ch \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=jglisse@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lvc-project@linuxtesting.org \
    --cc=n.zhandarovich@fintech.ru \
    --cc=stable@vger.kernel.org \
    /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