2023-05-22 05:08:17

by Hugh Dickins

Subject: [PATCH 05/31] mm/filemap: allow pte_offset_map_lock() to fail

filemap_map_pages() allow pte_offset_map_lock() to fail; and remove the
pmd_devmap_trans_unstable() check from filemap_map_pmd(), which can safely
return to filemap_map_pages() and let pte_offset_map_lock() discover that.

Signed-off-by: Hugh Dickins <[email protected]>
---
mm/filemap.c | 12 +++++-------
1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 28b42ee848a4..9e129ad43e0d 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -3408,13 +3408,6 @@ static bool filemap_map_pmd(struct vm_fault *vmf, struct folio *folio,
if (pmd_none(*vmf->pmd))
pmd_install(mm, vmf->pmd, &vmf->prealloc_pte);

- /* See comment in handle_pte_fault() */
- if (pmd_devmap_trans_unstable(vmf->pmd)) {
- folio_unlock(folio);
- folio_put(folio);
- return true;
- }
-
return false;
}

@@ -3501,6 +3494,11 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf,

addr = vma->vm_start + ((start_pgoff - vma->vm_pgoff) << PAGE_SHIFT);
vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, addr, &vmf->ptl);
+ if (!vmf->pte) {
+ folio_unlock(folio);
+ folio_put(folio);
+ goto out;
+ }
do {
again:
page = folio_file_page(folio, xas.xa_index);
--
2.35.3

2023-05-22 11:34:25

by Qi Zheng

Subject: Re: [PATCH 05/31] mm/filemap: allow pte_offset_map_lock() to fail

On 2023/5/22 12:54, Hugh Dickins wrote:
> filemap_map_pages() allow pte_offset_map_lock() to fail; and remove the
> pmd_devmap_trans_unstable() check from filemap_map_pmd(), which can safely
> return to filemap_map_pages() and let pte_offset_map_lock() discover that.
>
> Signed-off-by: Hugh Dickins <[email protected]>
> ---
> mm/filemap.c | 12 +++++-------
> 1 file changed, 5 insertions(+), 7 deletions(-)
>
> diff --git a/mm/filemap.c b/mm/filemap.c
> index 28b42ee848a4..9e129ad43e0d 100644
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -3408,13 +3408,6 @@ static bool filemap_map_pmd(struct vm_fault *vmf, struct folio *folio,
> if (pmd_none(*vmf->pmd))
> pmd_install(mm, vmf->pmd, &vmf->prealloc_pte);
>
> - /* See comment in handle_pte_fault() */
> - if (pmd_devmap_trans_unstable(vmf->pmd)) {
> - folio_unlock(folio);
> - folio_put(folio);
> - return true;
> - }
> -
> return false;
> }
>
> @@ -3501,6 +3494,11 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf,
>
> addr = vma->vm_start + ((start_pgoff - vma->vm_pgoff) << PAGE_SHIFT);
> vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, addr, &vmf->ptl);
> + if (!vmf->pte) {
> + folio_unlock(folio);
> + folio_put(folio);

Missing:
ret = VM_FAULT_NOPAGE;
?

> + goto out;
> + }
> do {
> again:
> page = folio_file_page(folio, xas.xa_index);

--
Thanks,
Qi

2023-05-24 02:40:29

by Hugh Dickins

Subject: Re: [PATCH 05/31] mm/filemap: allow pte_offset_map_lock() to fail

On Mon, 22 May 2023, Qi Zheng wrote:
> On 2023/5/22 12:54, Hugh Dickins wrote:
> > filemap_map_pages() allow pte_offset_map_lock() to fail; and remove the
> > pmd_devmap_trans_unstable() check from filemap_map_pmd(), which can safely
> > return to filemap_map_pages() and let pte_offset_map_lock() discover that.
> >
> > Signed-off-by: Hugh Dickins <[email protected]>
> > ---
> > mm/filemap.c | 12 +++++-------
> > 1 file changed, 5 insertions(+), 7 deletions(-)
> >
> > diff --git a/mm/filemap.c b/mm/filemap.c
> > index 28b42ee848a4..9e129ad43e0d 100644
> > --- a/mm/filemap.c
> > +++ b/mm/filemap.c
> > @@ -3408,13 +3408,6 @@ static bool filemap_map_pmd(struct vm_fault *vmf,
> > struct folio *folio,
> > if (pmd_none(*vmf->pmd))
> > pmd_install(mm, vmf->pmd, &vmf->prealloc_pte);
> > - /* See comment in handle_pte_fault() */
> > - if (pmd_devmap_trans_unstable(vmf->pmd)) {
> > - folio_unlock(folio);
> > - folio_put(folio);
> > - return true;
> > - }
> > -
> > return false;
> > }
> >
> > @@ -3501,6 +3494,11 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf,
> >
> > addr = vma->vm_start + ((start_pgoff - vma->vm_pgoff) << PAGE_SHIFT);
> > vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, addr, &vmf->ptl);
> > + if (!vmf->pte) {
> > + folio_unlock(folio);
> > + folio_put(folio);
>
> Missing:
> ret = VM_FAULT_NOPAGE;
> ?

No, not missed. Here ret is 0, which leads do_read_fault() to try
__do_fault() afterwards. Whereas VM_FAULT_NOPAGE would send it back
to userspace to retry the whole fault. Either will work, but I think
the intention of VM_FAULT_NOPAGE here in filemap_map_pages() is to say
"the page you want is now inserted", which is probably not the case.

Hugh

2023-05-24 04:01:46

by Qi Zheng

Subject: Re: [PATCH 05/31] mm/filemap: allow pte_offset_map_lock() to fail

On 2023/5/24 10:35, Hugh Dickins wrote:
> On Mon, 22 May 2023, Qi Zheng wrote:
>> On 2023/5/22 12:54, Hugh Dickins wrote:
>>> filemap_map_pages() allow pte_offset_map_lock() to fail; and remove the
>>> pmd_devmap_trans_unstable() check from filemap_map_pmd(), which can safely
>>> return to filemap_map_pages() and let pte_offset_map_lock() discover that.
>>>
>>> Signed-off-by: Hugh Dickins <[email protected]>
>>> ---
>>> mm/filemap.c | 12 +++++-------
>>> 1 file changed, 5 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/mm/filemap.c b/mm/filemap.c
>>> index 28b42ee848a4..9e129ad43e0d 100644
>>> --- a/mm/filemap.c
>>> +++ b/mm/filemap.c
>>> @@ -3408,13 +3408,6 @@ static bool filemap_map_pmd(struct vm_fault *vmf,
>>> struct folio *folio,
>>> if (pmd_none(*vmf->pmd))
>>> pmd_install(mm, vmf->pmd, &vmf->prealloc_pte);
>>> - /* See comment in handle_pte_fault() */
>>> - if (pmd_devmap_trans_unstable(vmf->pmd)) {
>>> - folio_unlock(folio);
>>> - folio_put(folio);
>>> - return true;
>>> - }
>>> -
>>> return false;
>>> }
>>>
>>> @@ -3501,6 +3494,11 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf,
>>>
>>> addr = vma->vm_start + ((start_pgoff - vma->vm_pgoff) << PAGE_SHIFT);
>>> vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, addr, &vmf->ptl);
>>> + if (!vmf->pte) {
>>> + folio_unlock(folio);
>>> + folio_put(folio);
>>
>> Missing:
>> ret = VM_FAULT_NOPAGE;
>> ?
>
> No, not missed. Here ret is 0, which leads do_read_fault() to try
> __do_fault() afterwards. Whereas VM_FAULT_NOPAGE would send it back
> to userspace to retry the whole fault. Either will work, but I think
> the intention of VM_FAULT_NOPAGE here in filemap_map_pages() is to say
> "the page you want is now inserted", which is probably not the case.

Got it. Thanks for the explanation.

>
> Hugh

--
Thanks,
Qi