Re: [PATCH 1/1] mm/mmu_gather: limit tlb batch count and add schedule point in tlb_batch_pages_flush

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



1. I try to increase the overcommit ratio of cpu to 1:2~1:3 in KVM
hypervisor, per-vm has the same number of vcpu with host cpu, then setup
2 or 3 vm.

2. Run ltpstress(20180926) test in per vm, both host and guest is non-preemptiable
kernel, vm dmesg will throw some rcu_sched warning.

3. PAGE_SIZE in my kernel config is 16K, tlb batch max count is 2015, it's too long in non-preemptible state.

The issue's orignal link:https://patchwork.kernel.org/project/linux-mm/patch/20220302013825.2290315-1-wangjianxing@xxxxxxxxxxx/


On 03/15/2022 08:55 PM, Jianxing Wang wrote:
free a large list of pages maybe cause rcu_sched starved on
non-preemptible kernels. howerver free_unref_page_list maybe can't
cond_resched as it maybe called in interrupt or atomic context,
especially can't detect atomic context in CONFIG_PREEMPTION=n.

tlb flush batch count depends on PAGE_SIZE, it's too large if
PAGE_SIZE > 4K, here limit max batch size with 4K.
And add schedule point in tlb_batch_pages_flush.

rcu: rcu_sched kthread starved for 5359 jiffies! g454793 f0x0
RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=19
[...]
Call Trace:
    free_unref_page_list+0x19c/0x270
    release_pages+0x3cc/0x498
    tlb_flush_mmu_free+0x44/0x70
    zap_pte_range+0x450/0x738
    unmap_page_range+0x108/0x240
    unmap_vmas+0x74/0xf0
    unmap_region+0xb0/0x120
    do_munmap+0x264/0x438
    vm_munmap+0x58/0xa0
    sys_munmap+0x10/0x20
    syscall_common+0x24/0x38

Signed-off-by: Jianxing Wang <wangjianxing@xxxxxxxxxxx>
---
  include/asm-generic/tlb.h | 7 ++++++-
  mm/mmu_gather.c           | 7 +++++--
  2 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h
index 2c68a545ffa7..47c7f93ca695 100644
--- a/include/asm-generic/tlb.h
+++ b/include/asm-generic/tlb.h
@@ -230,8 +230,13 @@ struct mmu_gather_batch {
  	struct page		*pages[0];
  };
+#if PAGE_SIZE > 4096UL
+#define MAX_GATHER_BATCH_SZ	4096
+#else
+#define MAX_GATHER_BATCH_SZ	PAGE_SIZE
+#endif
  #define MAX_GATHER_BATCH	\
-	((PAGE_SIZE - sizeof(struct mmu_gather_batch)) / sizeof(void *))
+	((MAX_GATHER_BATCH_SZ - sizeof(struct mmu_gather_batch)) / sizeof(void *))
/*
   * Limit the maximum number of mmu_gather batches to reduce a risk of soft
diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c
index afb7185ffdc4..f2c105810b3f 100644
--- a/mm/mmu_gather.c
+++ b/mm/mmu_gather.c
@@ -8,6 +8,7 @@
  #include <linux/rcupdate.h>
  #include <linux/smp.h>
  #include <linux/swap.h>
+#include <linux/slab.h>
#include <asm/pgalloc.h>
  #include <asm/tlb.h>
@@ -27,7 +28,7 @@ static bool tlb_next_batch(struct mmu_gather *tlb)
  	if (tlb->batch_count == MAX_GATHER_BATCH_COUNT)
  		return false;
- batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+	batch = kmalloc(MAX_GATHER_BATCH_SZ, GFP_NOWAIT | __GFP_NOWARN);
  	if (!batch)
  		return false;
@@ -49,6 +50,8 @@ static void tlb_batch_pages_flush(struct mmu_gather *tlb)
  	for (batch = &tlb->local; batch && batch->nr; batch = batch->next) {
  		free_pages_and_swap_cache(batch->pages, batch->nr);
  		batch->nr = 0;
+
+		cond_resched();
  	}
  	tlb->active = &tlb->local;
  }
@@ -59,7 +62,7 @@ static void tlb_batch_list_free(struct mmu_gather *tlb)
for (batch = tlb->local.next; batch; batch = next) {
  		next = batch->next;
-		free_pages((unsigned long)batch, 0);
+		kfree(batch);
  	}
  	tlb->local.next = NULL;
  }





[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux