Hi,
CC linux-fsdevel
On Mon, Dec 23, 2019 at 5:01 AM Chao Yu <yuchao0@xxxxxxxxxx> wrote:
> As Geert Uytterhoeven reported:
>
> for parameter HZ/50 in congestion_wait(BLK_RW_ASYNC, HZ/50);
>
> On some platforms, HZ can be less than 50, then unexpected 0 timeout
> jiffies will be set in congestion_wait().
>
> This patch introduces a macro DEFAULT_IO_TIMEOUT_JIFFIES to limit
> mininum value of timeout jiffies.
>
> Signed-off-by: Chao Yu <yuchao0@xxxxxxxxxx>
Thanks for your patch!
> ---
> fs/f2fs/compress.c | 3 ++-
> fs/f2fs/data.c | 5 +++--
> fs/f2fs/f2fs.h | 2 ++
> fs/f2fs/gc.c | 3 ++-
> fs/f2fs/inode.c | 3 ++-
> fs/f2fs/node.c | 3 ++-
> fs/f2fs/recovery.c | 6 ++++--
> fs/f2fs/segment.c | 12 ++++++++----
> fs/f2fs/super.c | 6 ++++--
> 9 files changed, 29 insertions(+), 14 deletions(-)
>
> diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c
> index 1bc86a54ad71..ee4fe8e644aa 100644
> --- a/fs/f2fs/compress.c
> +++ b/fs/f2fs/compress.c
> @@ -945,7 +945,8 @@ static int f2fs_write_raw_pages(struct compress_ctx *cc,
> } else if (ret == -EAGAIN) {
> ret = 0;
> cond_resched();
> - congestion_wait(BLK_RW_ASYNC, HZ/50);
> + congestion_wait(BLK_RW_ASYNC,
> + DEFAULT_IO_TIMEOUT_JIFFIES);
> lock_page(cc->rpages[i]);
> clear_page_dirty_for_io(cc->rpages[i]);
> goto retry_write;
> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> index f1f5c701228d..78b5c0b0287e 100644
> --- a/fs/f2fs/data.c
> +++ b/fs/f2fs/data.c
> @@ -2320,7 +2320,8 @@ int f2fs_encrypt_one_page(struct f2fs_io_info *fio)
> /* flush pending IOs and wait for a while in the ENOMEM case */
> if (PTR_ERR(fio->encrypted_page) == -ENOMEM) {
> f2fs_flush_merged_writes(fio->sbi);
> - congestion_wait(BLK_RW_ASYNC, HZ/50);
> + congestion_wait(BLK_RW_ASYNC,
> + DEFAULT_IO_TIMEOUT_JIFFIES);
> gfp_flags |= __GFP_NOFAIL;
> goto retry_encrypt;
> }
> @@ -2900,7 +2901,7 @@ static int f2fs_write_cache_pages(struct address_space *mapping,
> if (wbc->sync_mode == WB_SYNC_ALL) {
> cond_resched();
> congestion_wait(BLK_RW_ASYNC,
> - HZ/50);
> + DEFAULT_IO_TIMEOUT_JIFFIES);
> goto retry_write;
> }
> goto next;
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index 16edbf4e05e8..4bdc20a94185 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -559,6 +559,8 @@ enum {
>
> #define DEFAULT_RETRY_IO_COUNT 8 /* maximum retry read IO count */
>
> +#define DEFAULT_IO_TIMEOUT_JIFFIES (max_t(long, HZ/50, 1))
> +
> /* maximum retry quota flush count */
> #define DEFAULT_RETRY_QUOTA_FLUSH_COUNT 8
>
Seeing other file systems (ext4, xfs) and even core MM code suffers from
the same issue, perhaps it makes sense to move this into congestion_wait(),
i.e. increase the timeout to 1 if it's zero in the latter function?
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
