On Nov 30, 2020, at 8:49 PM, Chunguang Xu <brookxu.cn@xxxxxxxxx> wrote:
>
> From: Chunguang Xu <brookxu@xxxxxxxxxxx>
>
> patch cfd7323 introduces block bitmap prefetch, and expects to read
> block bitmaps of flex_bg through an IO. However, it seems to ignore
> the value range of s_log_groups_per_flex. In the scenario where the
> value of s_log_groups_per_flex is greater than 27, s_mb_prefetch or
> s_mb_prefetch_limit will overflow, cause a divide zero exception.
>
> In addition, the logic of calculating nr maybe also flawed, because
> the size of flexbg is fixed during a single mount, but s_mb_prefetch
> can be modified, which causes nr to fail to meet the value condition
> of [1, flexbg_size].
>
> PID: 3873 TASK: ffff88800f11d880 CPU: 2 COMMAND: "executor"
> #0 [ffff8880114a6ec0] __show_regs.cold.7 at ffffffff83cf29e2
> #1 [ffff8880114a6f40] do_trap at ffffffff81065c61
> #2 [ffff8880114a6f98] do_error_trap at ffffffff81065d65
> #3 [ffff8880114a6fe0] exc_divide_error at ffffffff83dd2fd4
> #4 [ffff8880114a7000] asm_exc_divide_error at ffffffff83e00872
> [exception RIP: ext4_mb_regular_allocator+3885]
> RIP: ffffffff8191258d RSP: ffff8880114a70b8 RFLAGS: 00010246
> RAX: 0000000000000000 RBX: 0000000000000000 RCX: ffffffff8191257a
> RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000005
> RBP: 0000000000000200 R8: ffff88800f11d880 R9: ffffed1001e23b11
> R10: ffff88800f11d887 R11: ffffed1001e23b10 R12: ffff888010147000
> R13: 0000000000000000 R14: 0000000000000002 R15: dffffc0000000000
> ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018
> #5 [ffff8880114a7260] ext4_mb_new_blocks at ffffffff8191b6ba
> #6 [ffff8880114a7420] ext4_new_meta_blocks at ffffffff81870d6f
> #7 [ffff8880114a74e8] ext4_xattr_block_set at ffffffff819ced37
> #8 [ffff8880114a7758] ext4_xattr_set_handle at ffffffff819d4776
> #9 [ffff8880114a7928] ext4_xattr_set at ffffffff819d501b
> RIP: 000000000045eb29 RSP: 00007ff74e97bc38 RFLAGS: 00000246
> RAX: ffffffffffffffda RBX: 000000000055bf00 RCX: 000000000045eb29
> RDX: 00000000200000c0 RSI: 0000000020000080 RDI: 0000000020000040
> RBP: 00000000004b068e R8: 0000000000000001 R9: 0000000000000000
> R10: 0000000000000002 R11: 0000000000000246 R12: 000000000055bf00
> R13: 00007fff50fc111f R14: 00007ff74e97bdc0 R15: 0000000000022000
> ORIG_RAX: 00000000000000bc CS: 0033 SS: 002b
>
> The maximum size of a single IO will be limited by multiple factors,
> such as max_hw_sectors, max_dev_sectors, BLK_DEF_MAX_SECTORS. The
> max_hw_sectors, max_dev_sectors are determined by the device, and
> BLK_DEF_MAX_SECTORS is a constant. In most scenarios, users will not
> modify max_sectors. Therefore, we can safely assume that the maximum
> size of a single IO is BLK_DEF_MAX_SECTORS. So far, we have determined
> the number of blocks that a single IO can hold. Usually the file
> system block is a multiple of the disk block, but we will ignore this
> for now. According to the current value of BLK_DEF_MAX_SECTORS and
BLK_DEF_MAX_SECTORS is 2560, or 1280KB, which isn't really a good
limit for the IO size. I think BLK_MAX_SEGMENT_SIZE = 65536 = 32MB
is a better upper limit to use in this case. This will almost always
be limited by he actual flexbg size, but provides a reasonable upper
limit that will still be useful into the future.
> comprehensive considerations, the maximum number of bitmap blocks that
> can be loaded by a single IO can be safely limited to 2^12. This maybe
> a good choice to solve divide zero problem and avoiding performance
> degradation.
>
> Reported-by: Tosk Robot <tencent_os_robot@xxxxxxxxxxx>
> Signed-off-by: Chunguang Xu <brookxu@xxxxxxxxxxx>
> Reviewed-by: Samuel Liao <samuelliao@xxxxxxxxxxx>
> ---
> fs/ext4/mballoc.c | 12 +++++++-----
> 1 file changed, 7 insertions(+), 5 deletions(-)
>
> diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
> index 24af9ed..06af4ca 100644
> --- a/fs/ext4/mballoc.c
> +++ b/fs/ext4/mballoc.c
> @@ -2395,9 +2395,10 @@ void ext4_mb_prefetch_fini(struct super_block *sb, ext4_group_t group,
>
> nr = sbi->s_mb_prefetch;
> if (ext4_has_feature_flex_bg(sb)) {
> - nr = (group / sbi->s_mb_prefetch) *
> - sbi->s_mb_prefetch;
> - nr = nr + sbi->s_mb_prefetch - group;
> + nr = 1 << sbi->s_log_groups_per_flex;
> + if (group & (nr - 1))
> + nr -= group & (nr - 1);
> + nr = min(nr, sbi->s_mb_prefetch);
> }
> prefetch_grp = ext4_mb_prefetch(sb, group,
> nr, &prefetch_ios);
> @@ -2700,7 +2701,7 @@ static int ext4_mb_init_backend(struct super_block *sb)
> ext4_group_t ngroups = ext4_get_groups_count(sb);
> ext4_group_t i;
> struct ext4_sb_info *sbi = EXT4_SB(sb);
> - int err;
> + int err, log;
(style) this variable should be declared inside the block below.
> struct ext4_group_desc *desc;
> struct ext4_group_info ***group_info;
> struct kmem_cache *cachep;
> @@ -2733,7 +2734,8 @@ static int ext4_mb_init_backend(struct super_block *sb)
>
> if (ext4_has_feature_flex_bg(sb)) {
> /* a single flex group is supposed to be read by a single IO */
> - sbi->s_mb_prefetch = 1 << sbi->s_es->s_log_groups_per_flex;
> + log = min_t(unsigned char, 12, sbi->s_es->s_log_groups_per_flex);
> + sbi->s_mb_prefetch = 1 << log;
Rather than hard-code "12" here, it would be better to use BLK_MAX_SEGMENT_SIZE
directly, so that it is clear in the future where this value came from, like:
if (ext4_has_feature_flex_bg(sb)) {
+ unsigned int len;
+
/* a single flex group is supposed to be read by a single IO */
- sbi->s_mb_prefetch = 1 << sbi->s_es->s_log_groups_per_flex;
+ len = min(BLK_MAX_SEGMENT_SIZE >> (sb->s_blocksize_bits - 9),
+ 1 << sbi->s_es->s_log_groups_per_flex);
+ sbi->s_mb_prefetch = len;
Note the need for "min_t()" can be avoided by using "1" or "1U" as needed in
the second half of "min()" to match the type of BLK_MAX_SEGMENT_SIZE.
Cheers, Andreas
Attachment:
signature.asc
Description: Message signed with OpenPGP
