On 10/15/26 04:20, [email protected] wrote: > From: Lu Yao <[email protected]> > > amdgpu_debugfs_gfxoff_count_read() stores a 64-bit GFXOFF entry count with > put_user(), but advances buf/pos/size by 4 bytes per iteration. > > Fixes: 0ad7347a64ac ("drm/amd: Add detailed GFXOFF stats to debugfs") > Signed-off-by: Lu Yao <[email protected]> > --- > drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 10 +++++----- > 1 file changed, 5 insertions(+), 5 deletions(-) > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c > b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c > index ab7ccfaea22a..aa44a3f0cdc4 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c > @@ -1451,7 +1451,7 @@ static ssize_t amdgpu_debugfs_gfxoff_count_read(struct > file *f, char __user *buf > ssize_t result = 0; > int r; > > - if (size & 0x3 || *pos & 0x3) > + if (size & 0x7 || *pos & 0x7) > return -EINVAL; > > r = pm_runtime_get_sync(adev_to_drm(adev)->dev); > @@ -1471,10 +1471,10 @@ static ssize_t > amdgpu_debugfs_gfxoff_count_read(struct file *f, char __user *buf > if (r) > goto out; > > - result += 4; > - buf += 4; > - *pos += 4; > - size -= 4; > + result += 8; > + buf += 8; > + *pos += 8; > + size -= 8;
That's a good catch, but the code is a bit questionable to begin with. Why are we reading the value multiple times and in binary instead of text? The same is true for the other amdgpu_debugfs_gfxoff_* functions. Regards, Christian. > } > > r = result;
