From: SeongJae Park <sj@kernel.org>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: SeongJae Park <sj@kernel.org>,
"Liam R. Howlett" <howlett@gmail.com>,
David Hildenbrand <david@redhat.com>,
Lorenzo Stoakes <lorenzo.stoakes@oracle.com>,
Shakeel Butt <shakeel.butt@linux.dev>,
Vlastimil Babka <vbabka@suse.cz>,
kernel-team@meta.com, linux-kernel@vger.kernel.org,
linux-mm@kvack.org
Subject: [PATCH 8/9] mm/madvise: batch tlb flushes for [process_]madvise(MADV_{DONTNEED[_LOCKED],FREE})
Date: Mon, 10 Mar 2025 10:23:17 -0700 [thread overview]
Message-ID: <20250310172318.653630-9-sj@kernel.org> (raw)
In-Reply-To: <20250310172318.653630-1-sj@kernel.org>
MADV_DONTNEED[_LOCKED] and MADV_FREE internal logics for
[process_]madvise() can be invoked with batched tlb flushes. Update
vector_madvise() and do_madvise(), which are called for the two system
calls respectively, to use those in the efficient way. Initialize an
mmu_gather object before starting the internal works, and flush the
gathered tlb entries at once after all the internal works are done.
Signed-off-by: SeongJae Park <sj@kernel.org>
---
mm/madvise.c | 51 +++++++++++++++++++++++++++++++++++++++++++++++----
1 file changed, 47 insertions(+), 4 deletions(-)
diff --git a/mm/madvise.c b/mm/madvise.c
index d7ea71c6422c..d5f4ce3041a4 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -905,6 +905,7 @@ static bool madvise_dontneed_free_valid_vma(struct vm_area_struct *vma,
struct madvise_behavior {
int behavior;
+ struct mmu_gather *tlb;
};
static long madvise_dontneed_free(struct vm_area_struct *vma,
@@ -964,9 +965,11 @@ static long madvise_dontneed_free(struct vm_area_struct *vma,
}
if (behavior == MADV_DONTNEED || behavior == MADV_DONTNEED_LOCKED)
- return madvise_dontneed_single_vma(NULL, vma, start, end);
+ return madvise_dontneed_single_vma(
+ madv_behavior->tlb, vma, start, end);
else if (behavior == MADV_FREE)
- return madvise_free_single_vma(NULL, vma, start, end);
+ return madvise_free_single_vma(
+ madv_behavior->tlb, vma, start, end);
else
return -EINVAL;
}
@@ -1639,6 +1642,32 @@ static void madvise_unlock(struct mm_struct *mm, int behavior)
mmap_read_unlock(mm);
}
+static bool madvise_batch_tlb_flush(int behavior)
+{
+ switch (behavior) {
+ case MADV_DONTNEED:
+ case MADV_DONTNEED_LOCKED:
+ return true;
+ default:
+ return false;
+ }
+}
+
+static void madvise_init_tlb(struct madvise_behavior *madv_behavior,
+ struct mm_struct *mm)
+{
+ if (!madvise_batch_tlb_flush(madv_behavior->behavior))
+ return;
+ tlb_gather_mmu(madv_behavior->tlb, mm);
+}
+
+static void madvise_finish_tlb(struct madvise_behavior *madv_behavior)
+{
+ if (!madvise_batch_tlb_flush(madv_behavior->behavior))
+ return;
+ tlb_finish_mmu(madv_behavior->tlb);
+}
+
static bool is_valid_madvise(unsigned long start, size_t len_in, int behavior)
{
size_t len;
@@ -1791,14 +1820,20 @@ static int madvise_do_behavior(struct mm_struct *mm,
int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior)
{
int error;
- struct madvise_behavior madv_behavior = {.behavior = behavior};
+ struct mmu_gather tlb;
+ struct madvise_behavior madv_behavior = {
+ .behavior = behavior,
+ .tlb = &tlb,
+ };
if (madvise_should_skip(start, len_in, behavior, &error))
return error;
error = madvise_lock(mm, behavior);
if (error)
return error;
+ madvise_init_tlb(&madv_behavior, mm);
error = madvise_do_behavior(mm, start, len_in, &madv_behavior);
+ madvise_finish_tlb(&madv_behavior);
madvise_unlock(mm, behavior);
return error;
@@ -1815,13 +1850,18 @@ static ssize_t vector_madvise(struct mm_struct *mm, struct iov_iter *iter,
{
ssize_t ret = 0;
size_t total_len;
- struct madvise_behavior madv_behavior = {.behavior = behavior};
+ struct mmu_gather tlb;
+ struct madvise_behavior madv_behavior = {
+ .behavior = behavior,
+ .tlb = &tlb,
+ };
total_len = iov_iter_count(iter);
ret = madvise_lock(mm, behavior);
if (ret)
return ret;
+ madvise_init_tlb(&madv_behavior, mm);
while (iov_iter_count(iter)) {
unsigned long start = (unsigned long)iter_iov_addr(iter);
@@ -1850,14 +1890,17 @@ static ssize_t vector_madvise(struct mm_struct *mm, struct iov_iter *iter,
}
/* Drop and reacquire lock to unwind race. */
+ madvise_finish_tlb(&madv_behavior);
madvise_unlock(mm, behavior);
madvise_lock(mm, behavior);
+ madvise_init_tlb(&madv_behavior, mm);
continue;
}
if (ret < 0)
break;
iov_iter_advance(iter, iter_iov_len(iter));
}
+ madvise_finish_tlb(&madv_behavior);
madvise_unlock(mm, behavior);
ret = (total_len - iov_iter_count(iter)) ? : ret;
--
2.39.5
next prev parent reply other threads:[~2025-03-10 17:23 UTC|newest]
Thread overview: 42+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-10 17:23 [PATCH 0/9] mm/madvise: batch tlb flushes for MADV_DONTNEED and MADV_FREE SeongJae Park
2025-03-10 17:23 ` [PATCH 1/9] mm/madvise: use is_memory_failure() from madvise_do_behavior() SeongJae Park
2025-03-11 11:27 ` Lorenzo Stoakes
2025-03-10 17:23 ` [PATCH 2/9] mm/madvise: split out populate behavior check logic SeongJae Park
2025-03-11 11:29 ` Lorenzo Stoakes
2025-03-10 17:23 ` [PATCH 3/9] mm/madvise: deduplicate madvise_do_behavior() skip case handlings SeongJae Park
2025-03-11 12:02 ` Lorenzo Stoakes
2025-03-11 20:54 ` SeongJae Park
2025-03-10 17:23 ` [PATCH 4/9] mm/madvise: remove len parameter of madvise_do_behavior() SeongJae Park
2025-03-11 12:05 ` Lorenzo Stoakes
2025-03-10 17:23 ` [PATCH 5/9] mm/madvise: define and use madvise_behavior struct for madvise_do_behavior() SeongJae Park
2025-03-11 12:17 ` Lorenzo Stoakes
2025-03-11 20:56 ` SeongJae Park
2025-03-12 5:47 ` Lorenzo Stoakes
2025-03-12 17:23 ` SeongJae Park
2025-03-10 17:23 ` [PATCH 6/9] mm/memory: split non-tlb flushing part from zap_page_range_single() SeongJae Park
2025-03-11 12:45 ` Lorenzo Stoakes
2025-03-11 20:58 ` SeongJae Park
2025-03-31 20:24 ` SeongJae Park
2025-04-01 1:45 ` Liam R. Howlett
2025-04-01 2:48 ` SeongJae Park
2025-04-01 14:03 ` Liam R. Howlett
2025-04-01 21:25 ` SeongJae Park
2025-03-10 17:23 ` [PATCH 7/9] mm/madvise: let madvise_{dontneed,free}_single_vma() caller batches tlb flushes SeongJae Park
2025-03-11 13:07 ` Lorenzo Stoakes
2025-03-11 21:00 ` SeongJae Park
2025-03-10 17:23 ` SeongJae Park [this message]
2025-03-11 13:59 ` [PATCH 8/9] mm/madvise: batch tlb flushes for [process_]madvise(MADV_{DONTNEED[_LOCKED],FREE}) Lorenzo Stoakes
2025-03-11 21:01 ` SeongJae Park
2025-04-01 21:17 ` SeongJae Park
2025-03-10 17:23 ` [PATCH 9/9] mm/madvise: remove !tlb support from madvise_{dontneed,free}_single_vma() SeongJae Park
2025-03-11 14:01 ` Lorenzo Stoakes
2025-03-11 21:02 ` SeongJae Park
2025-03-12 13:46 ` Lorenzo Stoakes
2025-04-01 21:22 ` SeongJae Park
2025-03-10 22:39 ` [PATCH 0/9] mm/madvise: batch tlb flushes for MADV_DONTNEED and MADV_FREE Andrew Morton
2025-03-10 23:15 ` Shakeel Butt
2025-03-10 23:36 ` Roman Gushchin
2025-03-11 11:17 ` Lorenzo Stoakes
2025-03-10 23:27 ` SeongJae Park
2025-03-11 12:49 ` Lorenzo Stoakes
2025-03-11 21:03 ` SeongJae Park
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=20250310172318.653630-9-sj@kernel.org \
--to=sj@kernel.org \
--cc=akpm@linux-foundation.org \
--cc=david@redhat.com \
--cc=howlett@gmail.com \
--cc=kernel-team@meta.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=lorenzo.stoakes@oracle.com \
--cc=shakeel.butt@linux.dev \
--cc=vbabka@suse.cz \
/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