Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 7 Jul 2021 13:00:39 +0000
From:   Dennis Zhou <dennis@kernel.org>
To:     Linus Torvalds <torvalds@linux-foundation.org>
Cc:     Tejun Heo <tj@kernel.org>, Christoph Lameter <cl@linux.com>,
        linux-mm@kvack.org, linux-kernel@vger.kernel.org
Subject: [GIT PULL] percpu fixes for v5.14-rc1
Message-ID: <YOWld9O5CZpzOUKA@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Precedence: bulk

Hi Linus,

This is just a single change to fix percpu depopulation. The code relied
on depopulation code written specifically for the free path and relied
on vmalloc to do the tlb flush lazily. As we're modifying the backing
pages during the lifetime of a chunk, we need to also flush the tlb
accordingly.

Guenter Roeck reported this issue in [1] on mips. I believe we just
happen to be lucky given the much larger chunk sizes on x86 and
consequently less churning of this memory.

[1] https://lore.kernel.org/lkml/20210702191140.GA3166599@roeck-us.net/

Thanks,
Dennis

The following changes since commit d6b63b5b7d7f363c6a54421533791e9849adf2e0:

  Merge tag 'sound-5.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound (2021-07-02 15:25:23 -0700)

are available in the Git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/dennis/percpu.git for-5.14-fixes

for you to fetch changes up to 93274f1dd6b0a615b299beddf99871fe81f91275:

  percpu: flush tlb in pcpu_reclaim_populated() (2021-07-04 18:30:17 +0000)

----------------------------------------------------------------
Dennis Zhou (1):
      percpu: flush tlb in pcpu_reclaim_populated()

 mm/percpu-km.c |  6 ++++++
 mm/percpu-vm.c |  5 +++--
 mm/percpu.c    | 32 ++++++++++++++++++++++++++------
 3 files changed, 35 insertions(+), 8 deletions(-)

diff --git a/mm/percpu-km.c b/mm/percpu-km.c
index c9d529dc7651..fe31aa19db81 100644
--- a/mm/percpu-km.c
+++ b/mm/percpu-km.c
@@ -32,6 +32,12 @@
 
 #include <linux/log2.h>
 
+static void pcpu_post_unmap_tlb_flush(struct pcpu_chunk *chunk,
+				      int page_start, int page_end)
+{
+	/* nothing */
+}
+
 static int pcpu_populate_chunk(struct pcpu_chunk *chunk,
 			       int page_start, int page_end, gfp_t gfp)
 {
diff --git a/mm/percpu-vm.c b/mm/percpu-vm.c
index ee5d89fcd66f..2054c9213c43 100644
--- a/mm/percpu-vm.c
+++ b/mm/percpu-vm.c
@@ -303,6 +303,9 @@ static int pcpu_populate_chunk(struct pcpu_chunk *chunk,
  * For each cpu, depopulate and unmap pages [@page_start,@page_end)
  * from @chunk.
  *
+ * Caller is required to call pcpu_post_unmap_tlb_flush() if not returning the
+ * region back to vmalloc() which will lazily flush the tlb.
+ *
  * CONTEXT:
  * pcpu_alloc_mutex.
  */
@@ -324,8 +327,6 @@ static void pcpu_depopulate_chunk(struct pcpu_chunk *chunk,
 
 	pcpu_unmap_pages(chunk, pages, page_start, page_end);
 
-	/* no need to flush tlb, vmalloc will handle it lazily */
-
 	pcpu_free_pages(chunk, pages, page_start, page_end);
 }
 
diff --git a/mm/percpu.c b/mm/percpu.c
index b4cebeca4c0c..7f2e0151c4e2 100644
--- a/mm/percpu.c
+++ b/mm/percpu.c
@@ -1572,6 +1572,7 @@ static void pcpu_chunk_depopulated(struct pcpu_chunk *chunk,
  *
  * pcpu_populate_chunk		- populate the specified range of a chunk
  * pcpu_depopulate_chunk	- depopulate the specified range of a chunk
+ * pcpu_post_unmap_tlb_flush	- flush tlb for the specified range of a chunk
  * pcpu_create_chunk		- create a new chunk
  * pcpu_destroy_chunk		- destroy a chunk, always preceded by full depop
  * pcpu_addr_to_page		- translate address to physical address
@@ -1581,6 +1582,8 @@ static int pcpu_populate_chunk(struct pcpu_chunk *chunk,
 			       int page_start, int page_end, gfp_t gfp);
 static void pcpu_depopulate_chunk(struct pcpu_chunk *chunk,
 				  int page_start, int page_end);
+static void pcpu_post_unmap_tlb_flush(struct pcpu_chunk *chunk,
+				      int page_start, int page_end);
 static struct pcpu_chunk *pcpu_create_chunk(gfp_t gfp);
 static void pcpu_destroy_chunk(struct pcpu_chunk *chunk);
 static struct page *pcpu_addr_to_page(void *addr);
@@ -2137,11 +2140,12 @@ static void pcpu_reclaim_populated(void)
 {
 	struct pcpu_chunk *chunk;
 	struct pcpu_block_md *block;
+	int freed_page_start, freed_page_end;
 	int i, end;
+	bool reintegrate;
 
 	lockdep_assert_held(&pcpu_lock);
 
-restart:
 	/*
 	 * Once a chunk is isolated to the to_depopulate list, the chunk is no
 	 * longer discoverable to allocations whom may populate pages.  The only
@@ -2157,6 +2161,9 @@ static void pcpu_reclaim_populated(void)
 		 * Scan chunk's pages in the reverse order to keep populated
 		 * pages close to the beginning of the chunk.
 		 */
+		freed_page_start = chunk->nr_pages;
+		freed_page_end = 0;
+		reintegrate = false;
 		for (i = chunk->nr_pages - 1, end = -1; i >= 0; i--) {
 			/* no more work to do */
 			if (chunk->nr_empty_pop_pages == 0)
@@ -2164,8 +2171,8 @@ static void pcpu_reclaim_populated(void)
 
 			/* reintegrate chunk to prevent atomic alloc failures */
 			if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_HIGH) {
-				pcpu_reintegrate_chunk(chunk);
-				goto restart;
+				reintegrate = true;
+				goto end_chunk;
 			}
 
 			/*
@@ -2194,16 +2201,29 @@ static void pcpu_reclaim_populated(void)
 			spin_lock_irq(&pcpu_lock);
 
 			pcpu_chunk_depopulated(chunk, i + 1, end + 1);
+			freed_page_start = min(freed_page_start, i + 1);
+			freed_page_end = max(freed_page_end, end + 1);
 
 			/* reset the range and continue */
 			end = -1;
 		}
 
-		if (chunk->free_bytes == pcpu_unit_size)
+end_chunk:
+		/* batch tlb flush per chunk to amortize cost */
+		if (freed_page_start < freed_page_end) {
+			spin_unlock_irq(&pcpu_lock);
+			pcpu_post_unmap_tlb_flush(chunk,
+						  freed_page_start,
+						  freed_page_end);
+			cond_resched();
+			spin_lock_irq(&pcpu_lock);
+		}
+
+		if (reintegrate || chunk->free_bytes == pcpu_unit_size)
 			pcpu_reintegrate_chunk(chunk);
 		else
-			list_move(&chunk->list,
-				  &pcpu_chunk_lists[pcpu_sidelined_slot]);
+			list_move_tail(&chunk->list,
+				       &pcpu_chunk_lists[pcpu_sidelined_slot]);
 	}
 }