From: David Hildenbrand <david@redhat.com>
To: linux-kernel@vger.kernel.org
Cc: linux-mm@kvack.org, linux-kselftest@vger.kernel.org,
Andrew Morton <akpm@linux-foundation.org>,
Shuah Khan <shuah@kernel.org>, Hugh Dickins <hughd@google.com>,
Vlastimil Babka <vbabka@suse.cz>, Peter Xu <peterx@redhat.com>,
Andrea Arcangeli <aarcange@redhat.com>,
"Matthew Wilcox (Oracle)" <willy@infradead.org>,
Jason Gunthorpe <jgg@nvidia.com>,
John Hubbard <jhubbard@nvidia.com>,
Janosch Frank <frankja@linux.ibm.com>
Subject: Re: [PATCH v1 6/7] mm/ksm: convert break_ksm() to use walk_page_range_vma()
Date: Thu, 20 Oct 2022 10:59:05 +0200 [thread overview]
Message-ID: <8c86678a-3bfb-3854-b1a9-ae5969e730b8@redhat.com> (raw)
In-Reply-To: <20220930141931.174362-7-david@redhat.com>
On 30.09.22 16:19, David Hildenbrand wrote:
> FOLL_MIGRATION exists only for the purpose of break_ksm(), and
> actually, there is not even the need to wait for the migration to
> finish, we only want to know if we're dealing with a KSM page.
>
> Using follow_page() just to identify a KSM page overcomplicates GUP
> code. Let's use walk_page_range_vma() instead, because we don't actually
> care about the page itself, we only need to know a single property --
> no need to even grab a reference on the page.
>
> In my setup (AMD Ryzen 9 3900X), running the KSM selftest to test unmerge
> performance on 2 GiB (taskset 0x8 ./ksm_tests -D -s 2048), this results in
> a performance degradation of ~4% (old: ~5010 MiB/s, new: ~4800 MiB/s).
> I don't think we particularly care for now.
>
> Signed-off-by: David Hildenbrand <david@redhat.com>
> ---
> mm/ksm.c | 70 +++++++++++++++++++++++++++++++++++++++++++++++++-------
> 1 file changed, 62 insertions(+), 8 deletions(-)
>
> diff --git a/mm/ksm.c b/mm/ksm.c
> index 4d7bcf7da7c3..814c1a37c323 100644
> --- a/mm/ksm.c
> +++ b/mm/ksm.c
> @@ -39,6 +39,7 @@
> #include <linux/freezer.h>
> #include <linux/oom.h>
> #include <linux/numa.h>
> +#include <linux/pagewalk.h>
>
> #include <asm/tlbflush.h>
> #include "internal.h"
> @@ -452,6 +453,60 @@ static inline bool ksm_test_exit(struct mm_struct *mm)
> return atomic_read(&mm->mm_users) == 0;
> }
>
> +int break_ksm_pud_entry(pud_t *pud, unsigned long addr, unsigned long next,
> + struct mm_walk *walk)
> +{
> + /* We only care about page tables to walk to a single base page. */
> + if (pud_leaf(*pud) || !pud_present(*pud))
> + return 1;
> + return 0;
> +}
> +
> +int break_ksm_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long next,
> + struct mm_walk *walk)
> +{
> + bool *ksm_page = walk->private;
> + struct page *page = NULL;
> + pte_t *pte, ptent;
> + spinlock_t *ptl;
> +
> + /* We only care about page tables to walk to a single base page. */
> + if (pmd_leaf(*pmd) || !pmd_present(*pmd))
> + return 1;
> +
> + /*
> + * We only lookup a single page (a) no need to iterate; and (b)
> + * always return 1 to exit immediately and not iterate in the caller.
> + */
> + pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
> + ptent = *pte;
> +
> + if (pte_none(ptent))
> + return 1;
As reported by Janosch, we fail to drop the lock here.
t480s: ~/git/linux ksm_unshare $ git diff
diff --git a/mm/ksm.c b/mm/ksm.c
index 26aec41b127c..94f8e114c89f 100644
--- a/mm/ksm.c
+++ b/mm/ksm.c
@@ -435,7 +435,7 @@ int break_ksm_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long next,
pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
if (pte_none(*pte))
- return 1;
+ goto out_unlock;
if (!pte_present(*pte)) {
swp_entry_t entry = pte_to_swp_entry(*pte);
@@ -451,6 +451,7 @@ int break_ksm_pmd_entry(pmd_t *pmd, unsigned long addr, unsigned long next,
}
if (page && PageKsm(page))
*ksm_page = true;
+out_unlock:
pte_unmap_unlock(pte, ptl);
return 1;
}
--
Thanks,
David / dhildenb
next prev parent reply other threads:[~2022-10-20 8:59 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-30 14:19 [PATCH v1 0/7] mm/ksm: break_ksm() cleanups and fixes David Hildenbrand
2022-09-30 14:19 ` [PATCH v1 1/7] selftests/vm: add test to measure MADV_UNMERGEABLE performance David Hildenbrand
2022-10-05 20:27 ` Peter Xu
2022-09-30 14:19 ` [PATCH v1 2/7] mm/ksm: simplify break_ksm() to not rely on VM_FAULT_WRITE David Hildenbrand
2022-10-05 20:29 ` Peter Xu
2022-09-30 14:19 ` [PATCH v1 3/7] mm: remove VM_FAULT_WRITE David Hildenbrand
2022-10-05 20:29 ` Peter Xu
2022-09-30 14:19 ` [PATCH v1 4/7] mm/ksm: fix KSM COW breaking with userfaultfd-wp via FAULT_FLAG_UNSHARE David Hildenbrand
2022-09-30 22:27 ` Andrew Morton
2022-10-01 8:13 ` David Hildenbrand
2022-10-05 20:35 ` Peter Xu
2022-10-06 9:29 ` David Hildenbrand
2022-10-06 19:04 ` Peter Xu
2022-10-20 10:05 ` David Hildenbrand
2022-09-30 14:19 ` [PATCH v1 5/7] mm/pagewalk: add walk_page_range_vma() David Hildenbrand
2022-10-05 20:42 ` Peter Xu
2022-10-06 9:35 ` David Hildenbrand
2022-09-30 14:19 ` [PATCH v1 6/7] mm/ksm: convert break_ksm() to use walk_page_range_vma() David Hildenbrand
2022-10-05 21:00 ` Peter Xu
2022-10-06 9:20 ` David Hildenbrand
2022-10-06 19:28 ` Peter Xu
2022-10-21 9:11 ` David Hildenbrand
2022-10-20 8:59 ` David Hildenbrand [this message]
2022-09-30 14:19 ` [PATCH v1 7/7] mm/gup: remove FOLL_MIGRATION David Hildenbrand
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=8c86678a-3bfb-3854-b1a9-ae5969e730b8@redhat.com \
--to=david@redhat.com \
--cc=aarcange@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=frankja@linux.ibm.com \
--cc=hughd@google.com \
--cc=jgg@nvidia.com \
--cc=jhubbard@nvidia.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-kselftest@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=peterx@redhat.com \
--cc=shuah@kernel.org \
--cc=vbabka@suse.cz \
--cc=willy@infradead.org \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).