From: Jason Gunthorpe <jgg@ziepe.ca>
To: Ralph Campbell <rcampbell@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>,
Felix.Kuehling@amd.com, linux-mm@kvack.org,
John Hubbard <jhubbard@nvidia.com>,
dri-devel@lists.freedesktop.org, amd-gfx@lists.freedesktop.org,
Christoph Hellwig <hch@lst.de>, Philip Yang <Philip.Yang@amd.com>
Subject: Re: [PATCH hmm 7/8] mm/hmm: return -EFAULT when setting HMM_PFN_ERROR on requested valid pages
Date: Thu, 12 Mar 2020 11:35:21 -0300 [thread overview]
Message-ID: <20200312143521.GN31668@ziepe.ca> (raw)
In-Reply-To: <910ae485-5400-3795-4353-ce4d47e3ae52@nvidia.com>
On Wed, Mar 11, 2020 at 06:36:47PM -0700, Ralph Campbell wrote:
> > @@ -390,8 +384,15 @@ static int hmm_vma_walk_pmd(pmd_t *pmdp,
> > return -EBUSY;
> > }
> > return hmm_pfns_fill(start, end, range, HMM_PFN_NONE);
> > - } else if (!pmd_present(pmd))
> > + }
> > +
> > + if (!pmd_present(pmd)) {
> > + hmm_range_need_fault(hmm_vma_walk, pfns, npages, 0, &fault,
> > + &write_fault);
> > + if (fault || write_fault)
> > + return -EFAULT;
> > return hmm_pfns_fill(start, end, range, HMM_PFN_ERROR);
>
> Shouldn't this fill with HMM_PFN_NONE instead of HMM_PFN_ERROR?
> Otherwise, when a THP is swapped out, you will get a different
> value than if a PTE is swapped out and you are prefetching/snapshotting.
If this is the case then the problem is that the return -EFAULT path
needs to do something else.. ie since the above code can't trigger
swap in, it is correct to return PFN_ERROR.
I'm completely guessing, but do we need to call pmd_to_swp_entry() and
handle things similarly to the pte? What swp_entries are valid for a
pmd?
Do you understand this better, or know how to trigger a !pmd_present
for test?
I suppose another option would be this:
if (!pmd_present(pmd)) {
hmm_range_need_fault(hmm_vma_walk, pfns, npages, 0, &fault,
&write_fault);
/* We can't handle this. Cause the PMD to be split and
* handle it in the pte handler. */
if (fault || write_fault)
return 0;
return hmm_pfns_fill(start, end, range, HMM_PFN_NONE);
}
Which, I think, must be correct, but inefficient?
Jason
WARNING: multiple messages have this Message-ID (diff)
From: Jason Gunthorpe <jgg@ziepe.ca>
To: Ralph Campbell <rcampbell@nvidia.com>
Cc: Philip Yang <Philip.Yang@amd.com>,
John Hubbard <jhubbard@nvidia.com>,
Felix.Kuehling@amd.com, amd-gfx@lists.freedesktop.org,
linux-mm@kvack.org, Jerome Glisse <jglisse@redhat.com>,
dri-devel@lists.freedesktop.org, Christoph Hellwig <hch@lst.de>
Subject: Re: [PATCH hmm 7/8] mm/hmm: return -EFAULT when setting HMM_PFN_ERROR on requested valid pages
Date: Thu, 12 Mar 2020 11:35:21 -0300 [thread overview]
Message-ID: <20200312143521.GN31668@ziepe.ca> (raw)
In-Reply-To: <910ae485-5400-3795-4353-ce4d47e3ae52@nvidia.com>
On Wed, Mar 11, 2020 at 06:36:47PM -0700, Ralph Campbell wrote:
> > @@ -390,8 +384,15 @@ static int hmm_vma_walk_pmd(pmd_t *pmdp,
> > return -EBUSY;
> > }
> > return hmm_pfns_fill(start, end, range, HMM_PFN_NONE);
> > - } else if (!pmd_present(pmd))
> > + }
> > +
> > + if (!pmd_present(pmd)) {
> > + hmm_range_need_fault(hmm_vma_walk, pfns, npages, 0, &fault,
> > + &write_fault);
> > + if (fault || write_fault)
> > + return -EFAULT;
> > return hmm_pfns_fill(start, end, range, HMM_PFN_ERROR);
>
> Shouldn't this fill with HMM_PFN_NONE instead of HMM_PFN_ERROR?
> Otherwise, when a THP is swapped out, you will get a different
> value than if a PTE is swapped out and you are prefetching/snapshotting.
If this is the case then the problem is that the return -EFAULT path
needs to do something else.. ie since the above code can't trigger
swap in, it is correct to return PFN_ERROR.
I'm completely guessing, but do we need to call pmd_to_swp_entry() and
handle things similarly to the pte? What swp_entries are valid for a
pmd?
Do you understand this better, or know how to trigger a !pmd_present
for test?
I suppose another option would be this:
if (!pmd_present(pmd)) {
hmm_range_need_fault(hmm_vma_walk, pfns, npages, 0, &fault,
&write_fault);
/* We can't handle this. Cause the PMD to be split and
* handle it in the pte handler. */
if (fault || write_fault)
return 0;
return hmm_pfns_fill(start, end, range, HMM_PFN_NONE);
}
Which, I think, must be correct, but inefficient?
Jason
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
WARNING: multiple messages have this Message-ID (diff)
From: Jason Gunthorpe <jgg@ziepe.ca>
To: Ralph Campbell <rcampbell@nvidia.com>
Cc: Philip Yang <Philip.Yang@amd.com>,
John Hubbard <jhubbard@nvidia.com>,
Felix.Kuehling@amd.com, amd-gfx@lists.freedesktop.org,
linux-mm@kvack.org, Jerome Glisse <jglisse@redhat.com>,
dri-devel@lists.freedesktop.org, Christoph Hellwig <hch@lst.de>
Subject: Re: [PATCH hmm 7/8] mm/hmm: return -EFAULT when setting HMM_PFN_ERROR on requested valid pages
Date: Thu, 12 Mar 2020 11:35:21 -0300 [thread overview]
Message-ID: <20200312143521.GN31668@ziepe.ca> (raw)
In-Reply-To: <910ae485-5400-3795-4353-ce4d47e3ae52@nvidia.com>
On Wed, Mar 11, 2020 at 06:36:47PM -0700, Ralph Campbell wrote:
> > @@ -390,8 +384,15 @@ static int hmm_vma_walk_pmd(pmd_t *pmdp,
> > return -EBUSY;
> > }
> > return hmm_pfns_fill(start, end, range, HMM_PFN_NONE);
> > - } else if (!pmd_present(pmd))
> > + }
> > +
> > + if (!pmd_present(pmd)) {
> > + hmm_range_need_fault(hmm_vma_walk, pfns, npages, 0, &fault,
> > + &write_fault);
> > + if (fault || write_fault)
> > + return -EFAULT;
> > return hmm_pfns_fill(start, end, range, HMM_PFN_ERROR);
>
> Shouldn't this fill with HMM_PFN_NONE instead of HMM_PFN_ERROR?
> Otherwise, when a THP is swapped out, you will get a different
> value than if a PTE is swapped out and you are prefetching/snapshotting.
If this is the case then the problem is that the return -EFAULT path
needs to do something else.. ie since the above code can't trigger
swap in, it is correct to return PFN_ERROR.
I'm completely guessing, but do we need to call pmd_to_swp_entry() and
handle things similarly to the pte? What swp_entries are valid for a
pmd?
Do you understand this better, or know how to trigger a !pmd_present
for test?
I suppose another option would be this:
if (!pmd_present(pmd)) {
hmm_range_need_fault(hmm_vma_walk, pfns, npages, 0, &fault,
&write_fault);
/* We can't handle this. Cause the PMD to be split and
* handle it in the pte handler. */
if (fault || write_fault)
return 0;
return hmm_pfns_fill(start, end, range, HMM_PFN_NONE);
}
Which, I think, must be correct, but inefficient?
Jason
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
next prev parent reply other threads:[~2020-03-12 14:35 UTC|newest]
Thread overview: 153+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-03-11 18:34 [PATCH hmm 0/8] Various error case bug fixes for hmm_range_fault() Jason Gunthorpe
2020-03-11 18:34 ` Jason Gunthorpe
2020-03-11 18:34 ` Jason Gunthorpe
2020-03-11 18:34 ` [PATCH hmm 1/8] mm/hmm: add missing unmaps of the ptep during hmm_vma_handle_pte() Jason Gunthorpe
2020-03-11 18:34 ` Jason Gunthorpe
2020-03-11 18:34 ` Jason Gunthorpe
2020-03-12 1:28 ` Ralph Campbell
2020-03-12 1:28 ` Ralph Campbell
2020-03-12 1:28 ` Ralph Campbell
2020-03-12 14:24 ` Jason Gunthorpe
2020-03-12 14:24 ` Jason Gunthorpe
2020-03-12 14:24 ` Jason Gunthorpe
2020-03-16 8:58 ` Christoph Hellwig
2020-03-16 8:58 ` Christoph Hellwig
2020-03-11 18:35 ` [PATCH hmm 2/8] mm/hmm: don't free the cached pgmap while scanning Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-12 1:29 ` Ralph Campbell
2020-03-12 1:29 ` Ralph Campbell
2020-03-12 1:29 ` Ralph Campbell
2020-03-16 9:02 ` Christoph Hellwig
2020-03-16 9:02 ` Christoph Hellwig
2020-03-16 18:07 ` Jason Gunthorpe
2020-03-16 18:07 ` Jason Gunthorpe
2020-03-16 18:07 ` Jason Gunthorpe
2020-03-16 18:13 ` Christoph Hellwig
2020-03-16 18:13 ` Christoph Hellwig
2020-03-16 19:23 ` Jason Gunthorpe
2020-03-16 19:23 ` Jason Gunthorpe
2020-03-16 19:23 ` Jason Gunthorpe
2020-03-11 18:35 ` [PATCH hmm 3/8] mm/hmm: do not call hmm_vma_walk_hole() while holding a spinlock Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-12 1:31 ` Ralph Campbell
2020-03-12 1:31 ` Ralph Campbell
2020-03-12 1:31 ` Ralph Campbell
2020-03-12 8:54 ` Steven Price
2020-03-12 8:54 ` Steven Price
2020-03-12 8:54 ` Steven Price
2020-03-12 10:28 ` [PATCH] mm/hmm: Simplify hmm_vma_walk_pud slightly Steven Price
2020-03-12 10:28 ` Steven Price
2020-03-12 10:28 ` Steven Price
2020-03-12 14:27 ` Jason Gunthorpe
2020-03-12 14:27 ` Jason Gunthorpe
2020-03-12 14:27 ` Jason Gunthorpe
2020-03-12 14:40 ` Steven Price
2020-03-12 14:40 ` Steven Price
2020-03-12 14:40 ` Steven Price
2020-03-12 15:11 ` Jason Gunthorpe
2020-03-12 15:11 ` Jason Gunthorpe
2020-03-12 15:11 ` Jason Gunthorpe
2020-03-12 16:16 ` Steven Price
2020-03-12 16:16 ` Steven Price
2020-03-12 16:16 ` Steven Price
2020-03-12 16:37 ` Jason Gunthorpe
2020-03-12 16:37 ` Jason Gunthorpe
2020-03-12 16:37 ` Jason Gunthorpe
2020-03-12 17:02 ` Steven Price
2020-03-12 17:02 ` Steven Price
2020-03-12 17:02 ` Steven Price
2020-03-12 17:17 ` Jason Gunthorpe
2020-03-12 17:17 ` Jason Gunthorpe
2020-03-12 17:17 ` Jason Gunthorpe
2020-03-13 19:55 ` Jason Gunthorpe
2020-03-13 19:55 ` Jason Gunthorpe
2020-03-13 19:55 ` Jason Gunthorpe
2020-03-13 21:04 ` Matthew Wilcox
2020-03-13 21:04 ` Matthew Wilcox
2020-03-13 21:04 ` Matthew Wilcox
2020-03-13 22:51 ` Jason Gunthorpe
2020-03-13 22:51 ` Jason Gunthorpe
2020-03-13 22:51 ` Jason Gunthorpe
2020-03-16 9:05 ` [PATCH hmm 3/8] mm/hmm: do not call hmm_vma_walk_hole() while holding a spinlock Christoph Hellwig
2020-03-16 9:05 ` Christoph Hellwig
2020-03-16 12:56 ` Jason Gunthorpe
2020-03-16 12:56 ` Jason Gunthorpe
2020-03-16 12:56 ` Jason Gunthorpe
2020-03-11 18:35 ` [PATCH hmm 4/8] mm/hmm: add missing pfns set to hmm_vma_walk_pmd() Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-12 1:33 ` Ralph Campbell
2020-03-12 1:33 ` Ralph Campbell
2020-03-12 1:33 ` Ralph Campbell
2020-03-16 9:06 ` Christoph Hellwig
2020-03-16 9:06 ` Christoph Hellwig
2020-03-11 18:35 ` [PATCH hmm 5/8] mm/hmm: add missing call to hmm_range_need_fault() before returning EFAULT Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-12 1:34 ` Ralph Campbell
2020-03-12 1:34 ` Ralph Campbell
2020-03-12 1:34 ` Ralph Campbell
2020-03-16 9:07 ` Christoph Hellwig
2020-03-16 9:07 ` Christoph Hellwig
2020-03-11 18:35 ` [PATCH hmm 6/8] mm/hmm: reorganize how !pte_present is handled in hmm_vma_handle_pte() Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-12 1:36 ` Ralph Campbell
2020-03-12 1:36 ` Ralph Campbell
2020-03-12 1:36 ` Ralph Campbell
2020-03-16 9:11 ` Christoph Hellwig
2020-03-16 9:11 ` Christoph Hellwig
2020-03-11 18:35 ` [PATCH hmm 7/8] mm/hmm: return -EFAULT when setting HMM_PFN_ERROR on requested valid pages Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-12 1:36 ` Ralph Campbell
2020-03-12 1:36 ` Ralph Campbell
2020-03-12 1:36 ` Ralph Campbell
2020-03-12 14:35 ` Jason Gunthorpe [this message]
2020-03-12 14:35 ` Jason Gunthorpe
2020-03-12 14:35 ` Jason Gunthorpe
2020-03-16 9:12 ` Christoph Hellwig
2020-03-16 9:12 ` Christoph Hellwig
2020-03-11 18:35 ` [PATCH hmm 8/8] mm/hmm: add missing call to hmm_pte_need_fault in HMM_PFN_SPECIAL handling Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-11 18:35 ` Jason Gunthorpe
2020-03-12 1:38 ` Ralph Campbell
2020-03-12 1:38 ` Ralph Campbell
2020-03-12 1:38 ` Ralph Campbell
2020-03-16 9:13 ` Christoph Hellwig
2020-03-16 9:13 ` Christoph Hellwig
2020-03-16 12:10 ` Jason Gunthorpe
2020-03-16 12:10 ` Jason Gunthorpe
2020-03-16 12:10 ` Jason Gunthorpe
2020-03-16 12:49 ` Christoph Hellwig
2020-03-16 12:49 ` Christoph Hellwig
2020-03-16 13:04 ` Jason Gunthorpe
2020-03-16 13:04 ` Jason Gunthorpe
2020-03-16 13:04 ` Jason Gunthorpe
2020-03-16 13:12 ` Christoph Hellwig
2020-03-16 13:12 ` Christoph Hellwig
2020-03-17 12:32 ` Christoph Hellwig
2020-03-17 12:32 ` Christoph Hellwig
2020-03-17 12:53 ` Jason Gunthorpe
2020-03-17 12:53 ` Jason Gunthorpe
2020-03-17 12:53 ` Jason Gunthorpe
2020-03-17 13:06 ` Christoph Hellwig
2020-03-17 13:06 ` Christoph Hellwig
2020-03-17 13:25 ` Jason Gunthorpe
2020-03-17 13:25 ` Jason Gunthorpe
2020-03-17 13:25 ` Jason Gunthorpe
2020-03-16 12:51 ` Christoph Hellwig
2020-03-16 12:51 ` Christoph Hellwig
2020-03-12 19:33 ` [PATCH hmm 9/8] mm/hmm: do not check pmd_protnone twice in hmm_vma_handle_pmd() Jason Gunthorpe
2020-03-12 19:33 ` Jason Gunthorpe
2020-03-12 19:33 ` Jason Gunthorpe
2020-03-12 23:50 ` Ralph Campbell
2020-03-12 23:50 ` Ralph Campbell
2020-03-12 23:50 ` Ralph Campbell
2020-03-16 9:14 ` Christoph Hellwig
2020-03-16 9:14 ` Christoph Hellwig
2020-03-16 18:25 ` [PATCH hmm 0/8] Various error case bug fixes for hmm_range_fault() Jason Gunthorpe
2020-03-16 18:25 ` Jason Gunthorpe
2020-03-16 18:25 ` Jason Gunthorpe
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=20200312143521.GN31668@ziepe.ca \
--to=jgg@ziepe.ca \
--cc=Felix.Kuehling@amd.com \
--cc=Philip.Yang@amd.com \
--cc=amd-gfx@lists.freedesktop.org \
--cc=dri-devel@lists.freedesktop.org \
--cc=hch@lst.de \
--cc=jglisse@redhat.com \
--cc=jhubbard@nvidia.com \
--cc=linux-mm@kvack.org \
--cc=rcampbell@nvidia.com \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.