2021-03-05 08:23:09

by Huang Jianan

Subject: [PATCH v4 2/2] erofs: decompress in endio if possible

z_erofs_decompressqueue_endio may not be executed in the atomic
context, for example, when dm-verity is turned on. In this scenario,
data can be decompressed directly to get rid of additional kworker
scheduling overhead. Also, it makes no sense to apply synchronous
decompression for such case.

Signed-off-by: Huang Jianan <[email protected]>
Signed-off-by: Guo Weichao <[email protected]>
---
fs/erofs/internal.h | 3 +++
fs/erofs/super.c | 4 ++++
fs/erofs/zdata.c | 16 +++++++++++++---
3 files changed, 20 insertions(+), 3 deletions(-)

diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index 67a7ec945686..b817cb85d67b 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -70,6 +70,9 @@ struct erofs_sb_info {

/* pseudo inode to manage cached pages */
struct inode *managed_cache;
+
+ /* decide whether to decompress synchronously */
+ bool readahead_sync_decompress;
#endif /* CONFIG_EROFS_FS_ZIP */
u32 blocks;
u32 meta_blkaddr;
diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index d5a6b9b888a5..77819efe9b15 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -175,6 +175,10 @@ static int erofs_read_superblock(struct super_block *sb)
sbi->root_nid = le16_to_cpu(dsb->root_nid);
sbi->inos = le64_to_cpu(dsb->inos);

+#ifdef CONFIG_EROFS_FS_ZIP
+ sbi->readahead_sync_decompress = false;
+#endif
+
sbi->build_time = le64_to_cpu(dsb->build_time);
sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);

diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index 6cb356c4217b..c21447c42eb0 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -706,9 +706,12 @@ static int z_erofs_do_read_page(struct z_erofs_decompress_frontend *fe,
goto out;
}

+static void z_erofs_decompressqueue_work(struct work_struct *work);
static void z_erofs_decompress_kickoff(struct z_erofs_decompressqueue *io,
bool sync, int bios)
{
+ struct erofs_sb_info *const sbi = EROFS_SB(io->sb);
+
/* wake up the caller thread for sync decompression */
if (sync) {
unsigned long flags;
@@ -720,8 +723,14 @@ static void z_erofs_decompress_kickoff(struct z_erofs_decompressqueue *io,
return;
}

- if (!atomic_add_return(bios, &io->pending_bios))
- queue_work(z_erofs_workqueue, &io->u.work);
+ if (!atomic_add_return(bios, &io->pending_bios)) {
+ if (in_atomic() || irqs_disabled()) {
+ queue_work(z_erofs_workqueue, &io->u.work);
+ sbi->readahead_sync_decompress = true;
+ } else {
+ z_erofs_decompressqueue_work(&io->u.work);
+ }
+ }
}

static bool z_erofs_page_is_invalidated(struct page *page)
@@ -1333,7 +1342,8 @@ static void z_erofs_readahead(struct readahead_control *rac)
struct erofs_sb_info *const sbi = EROFS_I_SB(inode);

unsigned int nr_pages = readahead_count(rac);
- bool sync = (nr_pages <= sbi->ctx.max_sync_decompress_pages);
+ bool sync = (sbi->readahead_sync_decompress &&
+ nr_pages <= sbi->ctx.max_sync_decompress_pages);
struct z_erofs_decompress_frontend f = DECOMPRESS_FRONTEND_INIT(inode);
struct page *page, *head = NULL;
LIST_HEAD(pagepool);
--
2.25.1

2021-03-05 09:22:13

by Gao Xiang

Subject: Re: [PATCH v4 2/2] erofs: decompress in endio if possible

On Fri, Mar 05, 2021 at 04:21:42PM +0800, Huang Jianan via Linux-erofs wrote:
> z_erofs_decompressqueue_endio may not be executed in the atomic
> context, for example, when dm-verity is turned on. In this scenario,
> data can be decompressed directly to get rid of additional kworker
> scheduling overhead. Also, it makes no sense to apply synchronous
> decompression for such case.

Could you resend the whole patchset [v4 RESEND] as a whole? Or you
can use --in-reply-to to attach the original version.

Yet I'd suggest you could resend them all this time...

>
> Signed-off-by: Huang Jianan <[email protected]>
> Signed-off-by: Guo Weichao <[email protected]>
> ---
> fs/erofs/internal.h | 3 +++
> fs/erofs/super.c | 4 ++++
> fs/erofs/zdata.c | 16 +++++++++++++---
> 3 files changed, 20 insertions(+), 3 deletions(-)
>
> diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
> index 67a7ec945686..b817cb85d67b 100644
> --- a/fs/erofs/internal.h
> +++ b/fs/erofs/internal.h
> @@ -70,6 +70,9 @@ struct erofs_sb_info {
>
> /* pseudo inode to manage cached pages */
> struct inode *managed_cache;
> +
> + /* decide whether to decompress synchronously */
> + bool readahead_sync_decompress;
> #endif /* CONFIG_EROFS_FS_ZIP */
> u32 blocks;
> u32 meta_blkaddr;
> diff --git a/fs/erofs/super.c b/fs/erofs/super.c
> index d5a6b9b888a5..77819efe9b15 100644
> --- a/fs/erofs/super.c
> +++ b/fs/erofs/super.c
> @@ -175,6 +175,10 @@ static int erofs_read_superblock(struct super_block *sb)
> sbi->root_nid = le16_to_cpu(dsb->root_nid);
> sbi->inos = le64_to_cpu(dsb->inos);
>
> +#ifdef CONFIG_EROFS_FS_ZIP
> + sbi->readahead_sync_decompress = false;
> +#endif

How about moving this stuff to erofs_default_options() as what we
did for max_sync_decompress_pages?

Thanks,
Gao Xiang