2021-04-07 20:34:45

by Huang Rui

Subject: Re: your mail

On Wed, Apr 07, 2021 at 09:27:46AM +0800, songqiang wrote:

Please add the description in the commit message and subject.

Thanks,
Ray

> Signed-off-by: songqiang <[email protected]>
> ---
> drivers/gpu/drm/ttm/ttm_page_alloc.c | 18 ++++++++++++++----
> 1 file changed, 14 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
> index 14660f723f71..f3698f0ad4d7 100644
> --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
> +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
> @@ -736,8 +736,16 @@ static void ttm_put_pages(struct page **pages, unsigned npages, int flags,
> if (++p != pages[i + j])
> break;
>
> - if (j == HPAGE_PMD_NR)
> + if (j == HPAGE_PMD_NR) {
> order = HPAGE_PMD_ORDER;
> + for (j = 1; j < HPAGE_PMD_NR; ++j)
> + page_ref_dec(pages[i+j]);
> + }
> }
> #endif
>
> @@ -868,10 +876,12 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags,
> p = alloc_pages(huge_flags, HPAGE_PMD_ORDER);
> if (!p)
> break;
> -
> - for (j = 0; j < HPAGE_PMD_NR; ++j)
> + for (j = 0; j < HPAGE_PMD_NR; ++j) {
> pages[i++] = p++;
> -
> + if (j > 0)
> + page_ref_inc(pages[i-1]);
> + }
> npages -= HPAGE_PMD_NR;
> }
> }
>
>
>
> _______________________________________________
> dri-devel mailing list
> [email protected]
> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Fdri-devel&amp;data=04%7C01%7Cray.huang%40amd.com%7C4ccc617b77d746db5af108d8f98db612%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637533734805563118%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&amp;sdata=9bSP90LYdJyJYJYmuphVmqk%2B3%2FE4JPrtXkQTbxwAt68%3D&amp;reserved=0

2021-04-07 20:41:08

by Christian König

Subject: Re: your mail

Thanks Ray for pointing this out. Looks like the mail ended up in my
spam folder otherwise.

Apart from that this patch is a really really big NAK. I can't count how
often I had to reject stuff like this!

Using the page reference for TTM pages is illegal and can lead to struct
page corruption.

Can you please describe why you need that?

Regards,
Christian.

Am 07.04.21 um 10:25 schrieb Huang Rui:
> On Wed, Apr 07, 2021 at 09:27:46AM +0800, songqiang wrote:
>
> Please add the description in the commit message and subject.
>
> Thanks,
> Ray
>
>> Signed-off-by: songqiang <[email protected]>
>> ---
>> drivers/gpu/drm/ttm/ttm_page_alloc.c | 18 ++++++++++++++----
>> 1 file changed, 14 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
>> index 14660f723f71..f3698f0ad4d7 100644
>> --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
>> +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
>> @@ -736,8 +736,16 @@ static void ttm_put_pages(struct page **pages, unsigned npages, int flags,
>> if (++p != pages[i + j])
>> break;
>>
>> - if (j == HPAGE_PMD_NR)
>> + if (j == HPAGE_PMD_NR) {
>> order = HPAGE_PMD_ORDER;
>> + for (j = 1; j < HPAGE_PMD_NR; ++j)
>> + page_ref_dec(pages[i+j]);
>> + }
>> }
>> #endif
>>
>> @@ -868,10 +876,12 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags,
>> p = alloc_pages(huge_flags, HPAGE_PMD_ORDER);
>> if (!p)
>> break;
>> -
>> - for (j = 0; j < HPAGE_PMD_NR; ++j)
>> + for (j = 0; j < HPAGE_PMD_NR; ++j) {
>> pages[i++] = p++;
>> -
>> + if (j > 0)
>> + page_ref_inc(pages[i-1]);
>> + }
>> npages -= HPAGE_PMD_NR;
>> }
>> }
>>
>>
>>
>> _______________________________________________
>> dri-devel mailing list
>> [email protected]
>> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Fdri-devel&amp;data=04%7C01%7Cray.huang%40amd.com%7C4ccc617b77d746db5af108d8f98db612%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637533734805563118%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&amp;sdata=9bSP90LYdJyJYJYmuphVmqk%2B3%2FE4JPrtXkQTbxwAt68%3D&amp;reserved=0