* [PATCH] erofs-utils: no compression case for tail-end block in vle_write_indexes()
@ 2021-07-23 3:49 Yue Hu
2021-07-23 5:02 ` Gao Xiang
0 siblings, 1 reply; 3+ messages in thread
From: Yue Hu @ 2021-07-23 3:49 UTC (permalink / raw)
To: linux-erofs, xiang; +Cc: huyue2, yuchao0, zbestahu
From: Yue Hu <huyue2@yulong.com>
Note that count value will be always greater than EROFS_BLKSIZ when
calling erofs_compress_destsize() in vle_compress_one(). So, the d1
always >= 1 for compressed block in vle_write_indexes(). That is to
say tail-end block can't be compressed.
Signed-off-by: Yue Hu <huyue2@yulong.com>
---
lib/compress.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/lib/compress.c b/lib/compress.c
index af0c720..93fc543 100644
--- a/lib/compress.c
+++ b/lib/compress.c
@@ -73,10 +73,9 @@ static void vle_write_indexes(struct z_erofs_vle_compress_ctx *ctx,
di.di_clusterofs = cpu_to_le16(ctx->clusterofs);
- /* whether the tail-end (un)compressed block or not */
+ /* whether the tail-end uncompressed block or not */
if (!d1) {
- type = raw ? Z_EROFS_VLE_CLUSTER_TYPE_PLAIN :
- Z_EROFS_VLE_CLUSTER_TYPE_HEAD;
+ type = Z_EROFS_VLE_CLUSTER_TYPE_PLAIN;
advise = cpu_to_le16(type << Z_EROFS_VLE_DI_CLUSTER_TYPE_BIT);
di.di_advise = advise;
--
1.9.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] erofs-utils: no compression case for tail-end block in vle_write_indexes()
2021-07-23 3:49 [PATCH] erofs-utils: no compression case for tail-end block in vle_write_indexes() Yue Hu
@ 2021-07-23 5:02 ` Gao Xiang
2021-07-23 6:10 ` Yue Hu
0 siblings, 1 reply; 3+ messages in thread
From: Gao Xiang @ 2021-07-23 5:02 UTC (permalink / raw)
To: Yue Hu; +Cc: yuchao0, zbestahu, huyue2, xiang, linux-erofs
Hi Yue,
On Fri, Jul 23, 2021 at 11:49:45AM +0800, Yue Hu wrote:
> From: Yue Hu <huyue2@yulong.com>
>
> Note that count value will be always greater than EROFS_BLKSIZ when
> calling erofs_compress_destsize() in vle_compress_one(). So, the d1
> always >= 1 for compressed block in vle_write_indexes(). That is to
> say tail-end block can't be compressed.
>
> Signed-off-by: Yue Hu <huyue2@yulong.com>
I once thought tail end block can be compressed with < EROFS_BLKSIZ as
well (as long as it has some benefit) and also tail-packing inline
like uncompressed cases together after compress extents, so that when we
reading the last compress extent meta block, the tail-packing compressed
data can be loaded together.
Currently, I think we could do that, but could you add some DBG_BUGON
like DBG_BUGON(!raw) right above
"type = Z_EROFS_VLE_CLUSTER_TYPE_PLAIN;" and a TODO comment to mention
tail-packing inline compressed cases is under TODO list? That would be
much better.
Thanks,
Gao Xiang
> ---
> lib/compress.c | 5 ++---
> 1 file changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/lib/compress.c b/lib/compress.c
> index af0c720..93fc543 100644
> --- a/lib/compress.c
> +++ b/lib/compress.c
> @@ -73,10 +73,9 @@ static void vle_write_indexes(struct z_erofs_vle_compress_ctx *ctx,
>
> di.di_clusterofs = cpu_to_le16(ctx->clusterofs);
>
> - /* whether the tail-end (un)compressed block or not */
> + /* whether the tail-end uncompressed block or not */
> if (!d1) {
> - type = raw ? Z_EROFS_VLE_CLUSTER_TYPE_PLAIN :
> - Z_EROFS_VLE_CLUSTER_TYPE_HEAD;
> + type = Z_EROFS_VLE_CLUSTER_TYPE_PLAIN;
> advise = cpu_to_le16(type << Z_EROFS_VLE_DI_CLUSTER_TYPE_BIT);
>
> di.di_advise = advise;
> --
> 1.9.1
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] erofs-utils: no compression case for tail-end block in vle_write_indexes()
2021-07-23 5:02 ` Gao Xiang
@ 2021-07-23 6:10 ` Yue Hu
0 siblings, 0 replies; 3+ messages in thread
From: Yue Hu @ 2021-07-23 6:10 UTC (permalink / raw)
To: Gao Xiang; +Cc: yuchao0, zbestahu, huyue2, xiang, linux-erofs
On Fri, 23 Jul 2021 13:02:14 +0800
Gao Xiang <hsiangkao@linux.alibaba.com> wrote:
> Hi Yue,
>
> On Fri, Jul 23, 2021 at 11:49:45AM +0800, Yue Hu wrote:
> > From: Yue Hu <huyue2@yulong.com>
> >
> > Note that count value will be always greater than EROFS_BLKSIZ when
> > calling erofs_compress_destsize() in vle_compress_one(). So, the d1
> > always >= 1 for compressed block in vle_write_indexes(). That is to
> > say tail-end block can't be compressed.
> >
> > Signed-off-by: Yue Hu <huyue2@yulong.com>
>
> I once thought tail end block can be compressed with < EROFS_BLKSIZ as
> well (as long as it has some benefit) and also tail-packing inline
> like uncompressed cases together after compress extents, so that when we
> reading the last compress extent meta block, the tail-packing compressed
> data can be loaded together.
Understood.
>
> Currently, I think we could do that, but could you add some DBG_BUGON
> like DBG_BUGON(!raw) right above
> "type = Z_EROFS_VLE_CLUSTER_TYPE_PLAIN;" and a TODO comment to mention
> tail-packing inline compressed cases is under TODO list? That would be
Will do it in v2.
Thanks.
> much better.
>
> Thanks,
> Gao Xiang
>
> > ---
> > lib/compress.c | 5 ++---
> > 1 file changed, 2 insertions(+), 3 deletions(-)
> >
> > diff --git a/lib/compress.c b/lib/compress.c
> > index af0c720..93fc543 100644
> > --- a/lib/compress.c
> > +++ b/lib/compress.c
> > @@ -73,10 +73,9 @@ static void vle_write_indexes(struct z_erofs_vle_compress_ctx *ctx,
> >
> > di.di_clusterofs = cpu_to_le16(ctx->clusterofs);
> >
> > - /* whether the tail-end (un)compressed block or not */
> > + /* whether the tail-end uncompressed block or not */
> > if (!d1) {
> > - type = raw ? Z_EROFS_VLE_CLUSTER_TYPE_PLAIN :
> > - Z_EROFS_VLE_CLUSTER_TYPE_HEAD;
> > + type = Z_EROFS_VLE_CLUSTER_TYPE_PLAIN;
> > advise = cpu_to_le16(type << Z_EROFS_VLE_DI_CLUSTER_TYPE_BIT);
> >
> > di.di_advise = advise;
> > --
> > 1.9.1
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2021-07-23 6:10 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-07-23 3:49 [PATCH] erofs-utils: no compression case for tail-end block in vle_write_indexes() Yue Hu
2021-07-23 5:02 ` Gao Xiang
2021-07-23 6:10 ` Yue Hu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).