Re: [PATCH v2] mm: memdup_user*() should use same gfp flags

[Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>]

On 2021/01/22 10:35, Andrew Morton wrote:
> On Wed, 20 Jan 2021 19:34:36 +0900 Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> wrote:
> 
>> syzbot is reporting that memdup_user_nul() which receives user-controlled
>> size (which can be up to (INT_MAX & PAGE_MASK)) via vfs_write() will hit
>> order >= MAX_ORDER path [1].
>>
>> Let's add __GFP_NOWARN to memdup_user_nul() as with commit 6c8fcc096be9d02f
>> ("mm: don't let userspace spam allocations warnings"). Also use GFP_USER as
>> with commit 6c2c97a24f096e32 ("memdup_user(): switch to GFP_USER").
> 
> That commit failed to explain why a switch to GFP_USER was performed,
> so that commit isn't a good substitute for an explanation of this
> change.

For example, commit 2f77d107050abc14 ("Fix incorrect user space access locking
in mincore()") silently converted GFP_KERNEL to GFP_USER.

  #define GFP_KERNEL	(__GFP_RECLAIM | __GFP_IO | __GFP_FS)
  #define GFP_USER	(__GFP_RECLAIM | __GFP_IO | __GFP_FS | __GFP_HARDWALL)

 * %GFP_KERNEL is typical for kernel-internal allocations. The caller requires
 * %ZONE_NORMAL or a lower zone for direct access but can direct reclaim.

 * %GFP_USER is for userspace allocations that also need to be directly
 * accessibly by the kernel or hardware. It is typically used by hardware
 * for buffers that are mapped to userspace (e.g. graphics) that hardware
 * still must DMA to. cpuset limits are enforced for these allocations.

 * %__GFP_HARDWALL enforces the cpuset memory allocation policy.

> 
> So...  please fully describe the reason for this change right here in
> this patch's changelog.

I guess that GFP_USER is chosen by cautious developers when memory is
allocated by userspace request. Is there a guideline for when to use GFP_USER ?