FAILED: patch "[PATCH] erofs: fix lz4 inplace decompression" failed to apply to 5.10-stable tree

FAILED: patch "[PATCH] erofs: fix lz4 inplace decompression" failed to apply to 5.10-stable tree

[gregkh]

The patch below does not apply to the 5.10-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable@vger.kernel.org>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-5.10.y
git checkout FETCH_HEAD
git cherry-pick -x 3c12466b6b7bf1e56f9b32c366a3d83d87afb4de
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<stable@vger.kernel.org>' --in-reply-to '2024012650-altitude-gush-572f@gregkh' --subject-prefix 'PATCH 5.10.y' HEAD^..

Possible dependencies:

3c12466b6b7b ("erofs: fix lz4 inplace decompression")
123ec246ebe3 ("erofs: get rid of the remaining kmap_atomic()")
ab749badf9f4 ("erofs: support unaligned data decompression")
10e5f6e482e1 ("erofs: introduce z_erofs_fixup_insize")
d67aee76d418 ("erofs: tidy up z_erofs_lz4_decompress")
7e508f2ca8bb ("erofs: rename lz4_0pading to zero_padding")
eaa9172ad988 ("erofs: get rid of ->lru usage")
622ceaddb764 ("erofs: lzma compression support")
966edfb0a3dc ("erofs: rename some generic methods in decompressor")
386292919c25 ("erofs: introduce readmore decompression strategy")
8f89926290c4 ("erofs: get compression algorithms directly on mapping")
dfeab2e95a75 ("erofs: add multiple device support")
e62424651f43 ("erofs: decouple basic mount options from fs_context")
5b6e7e120e71 ("erofs: remove the fast path of per-CPU buffer decompression")
2e5fd489a4e5 ("Merge tag 'libnvdimm-for-5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm")

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 3c12466b6b7bf1e56f9b32c366a3d83d87afb4de Mon Sep 17 00:00:00 2001
From: Gao Xiang <xiang@kernel.org>
Date: Wed, 6 Dec 2023 12:55:34 +0800
Subject: [PATCH] erofs: fix lz4 inplace decompression

Currently EROFS can map another compressed buffer for inplace
decompression, that was used to handle the cases that some pages of
compressed data are actually not in-place I/O.

However, like most simple LZ77 algorithms, LZ4 expects the compressed
data is arranged at the end of the decompressed buffer and it
explicitly uses memmove() to handle overlapping:
  __________________________________________________________
 |_ direction of decompression --> ____ |_ compressed data _|

Although EROFS arranges compressed data like this, it typically maps two
individual virtual buffers so the relative order is uncertain.
Previously, it was hardly observed since LZ4 only uses memmove() for
short overlapped literals and x86/arm64 memmove implementations seem to
completely cover it up and they don't have this issue.  Juhyung reported
that EROFS data corruption can be found on a new Intel x86 processor.
After some analysis, it seems that recent x86 processors with the new
FSRM feature expose this issue with "rep movsb".

Let's strictly use the decompressed buffer for lz4 inplace
decompression for now.  Later, as an useful improvement, we could try
to tie up these two buffers together in the correct order.

Reported-and-tested-by: Juhyung Park <qkrwngud825@gmail.com>
Closes: https://lore.kernel.org/r/CAD14+f2AVKf8Fa2OO1aAUdDNTDsVzzR6ctU_oJSmTyd6zSYR2Q@mail.gmail.com
Fixes: 0ffd71bcc3a0 ("staging: erofs: introduce LZ4 decompression inplace")
Fixes: 598162d05080 ("erofs: support decompress big pcluster for lz4 backend")
Cc: stable <stable@vger.kernel.org> # 5.4+
Tested-by: Yifan Zhao <zhaoyifan@sjtu.edu.cn>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Link: https://lore.kernel.org/r/20231206045534.3920847-1-hsiangkao@linux.alibaba.com

diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index 021be5feb1bc..e0d609c3958f 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -121,11 +121,11 @@ static int z_erofs_lz4_prepare_dstpages(struct z_erofs_lz4_decompress_ctx *ctx,
 }
 
 static void *z_erofs_lz4_handle_overlap(struct z_erofs_lz4_decompress_ctx *ctx,
-			void *inpage, unsigned int *inputmargin, int *maptype,
-			bool may_inplace)
+			void *inpage, void *out, unsigned int *inputmargin,
+			int *maptype, bool may_inplace)
 {
 	struct z_erofs_decompress_req *rq = ctx->rq;
-	unsigned int omargin, total, i, j;
+	unsigned int omargin, total, i;
 	struct page **in;
 	void *src, *tmp;
 
@@ -135,12 +135,13 @@ static void *z_erofs_lz4_handle_overlap(struct z_erofs_lz4_decompress_ctx *ctx,
 		    omargin < LZ4_DECOMPRESS_INPLACE_MARGIN(rq->inputsize))
 			goto docopy;
 
-		for (i = 0; i < ctx->inpages; ++i) {
-			DBG_BUGON(rq->in[i] == NULL);
-			for (j = 0; j < ctx->outpages - ctx->inpages + i; ++j)
-				if (rq->out[j] == rq->in[i])
-					goto docopy;
-		}
+		for (i = 0; i < ctx->inpages; ++i)
+			if (rq->out[ctx->outpages - ctx->inpages + i] !=
+			    rq->in[i])
+				goto docopy;
+		kunmap_local(inpage);
+		*maptype = 3;
+		return out + ((ctx->outpages - ctx->inpages) << PAGE_SHIFT);
 	}
 
 	if (ctx->inpages <= 1) {
@@ -148,7 +149,6 @@ static void *z_erofs_lz4_handle_overlap(struct z_erofs_lz4_decompress_ctx *ctx,
 		return inpage;
 	}
 	kunmap_local(inpage);
-	might_sleep();
 	src = erofs_vm_map_ram(rq->in, ctx->inpages);
 	if (!src)
 		return ERR_PTR(-ENOMEM);
@@ -204,12 +204,12 @@ int z_erofs_fixup_insize(struct z_erofs_decompress_req *rq, const char *padbuf,
 }
 
 static int z_erofs_lz4_decompress_mem(struct z_erofs_lz4_decompress_ctx *ctx,
-				      u8 *out)
+				      u8 *dst)
 {
 	struct z_erofs_decompress_req *rq = ctx->rq;
 	bool support_0padding = false, may_inplace = false;
 	unsigned int inputmargin;
-	u8 *headpage, *src;
+	u8 *out, *headpage, *src;
 	int ret, maptype;
 
 	DBG_BUGON(*rq->in == NULL);
@@ -230,11 +230,12 @@ static int z_erofs_lz4_decompress_mem(struct z_erofs_lz4_decompress_ctx *ctx,
 	}
 
 	inputmargin = rq->pageofs_in;
-	src = z_erofs_lz4_handle_overlap(ctx, headpage, &inputmargin,
+	src = z_erofs_lz4_handle_overlap(ctx, headpage, dst, &inputmargin,
 					 &maptype, may_inplace);
 	if (IS_ERR(src))
 		return PTR_ERR(src);
 
+	out = dst + rq->pageofs_out;
 	/* legacy format could compress extra data in a pcluster. */
 	if (rq->partial_decoding || !support_0padding)
 		ret = LZ4_decompress_safe_partial(src + inputmargin, out,
@@ -265,7 +266,7 @@ static int z_erofs_lz4_decompress_mem(struct z_erofs_lz4_decompress_ctx *ctx,
 		vm_unmap_ram(src, ctx->inpages);
 	} else if (maptype == 2) {
 		erofs_put_pcpubuf(src);
-	} else {
+	} else if (maptype != 3) {
 		DBG_BUGON(1);
 		return -EFAULT;
 	}
@@ -308,7 +309,7 @@ static int z_erofs_lz4_decompress(struct z_erofs_decompress_req *rq,
 	}
 
 dstmap_out:
-	ret = z_erofs_lz4_decompress_mem(&ctx, dst + rq->pageofs_out);
+	ret = z_erofs_lz4_decompress_mem(&ctx, dst);
 	if (!dst_maptype)
 		kunmap_local(dst);
 	else if (dst_maptype == 2)

[PATCH 5.10.y] erofs: fix lz4 inplace decompression

[Gao Xiang]

commit 3c12466b6b7bf1e56f9b32c366a3d83d87afb4de upstream.

Currently EROFS can map another compressed buffer for inplace
decompression, that was used to handle the cases that some pages of
compressed data are actually not in-place I/O.

However, like most simple LZ77 algorithms, LZ4 expects the compressed
data is arranged at the end of the decompressed buffer and it
explicitly uses memmove() to handle overlapping:
  __________________________________________________________
 |_ direction of decompression --> ____ |_ compressed data _|

Although EROFS arranges compressed data like this, it typically maps two
individual virtual buffers so the relative order is uncertain.
Previously, it was hardly observed since LZ4 only uses memmove() for
short overlapped literals and x86/arm64 memmove implementations seem to
completely cover it up and they don't have this issue.  Juhyung reported
that EROFS data corruption can be found on a new Intel x86 processor.
After some analysis, it seems that recent x86 processors with the new
FSRM feature expose this issue with "rep movsb".

Let's strictly use the decompressed buffer for lz4 inplace
decompression for now.  Later, as an useful improvement, we could try
to tie up these two buffers together in the correct order.

Reported-and-tested-by: Juhyung Park <qkrwngud825@gmail.com>
Closes: https://lore.kernel.org/r/CAD14+f2AVKf8Fa2OO1aAUdDNTDsVzzR6ctU_oJSmTyd6zSYR2Q@mail.gmail.com
Fixes: 0ffd71bcc3a0 ("staging: erofs: introduce LZ4 decompression inplace")
Fixes: 598162d05080 ("erofs: support decompress big pcluster for lz4 backend")
Cc: stable <stable@vger.kernel.org> # 5.4+
Tested-by: Yifan Zhao <zhaoyifan@sjtu.edu.cn>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Link: https://lore.kernel.org/r/20231206045534.3920847-1-hsiangkao@linux.alibaba.com
---
Adapt 5.10.y codebase due to non-trivial conflicts out of
recent new features & cleanups.

I've done my own EROFS tests, yet more test on new x86
platform triggered by FSRM is helpful too.

( This fix for 5.4.y codebase is still ongoing, and I will
  post later. )

 fs/erofs/decompressor.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)

diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index f921580b56cb..36693924db18 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -24,7 +24,8 @@ struct z_erofs_decompressor {
 	 */
 	int (*prepare_destpages)(struct z_erofs_decompress_req *rq,
 				 struct list_head *pagepool);
-	int (*decompress)(struct z_erofs_decompress_req *rq, u8 *out);
+	int (*decompress)(struct z_erofs_decompress_req *rq, u8 *out,
+			  u8 *obase);
 	char *name;
 };
 
@@ -114,10 +115,13 @@ static void *generic_copy_inplace_data(struct z_erofs_decompress_req *rq,
 	return tmp;
 }
 
-static int z_erofs_lz4_decompress(struct z_erofs_decompress_req *rq, u8 *out)
+static int z_erofs_lz4_decompress(struct z_erofs_decompress_req *rq, u8 *out,
+				  u8 *obase)
 {
+	const uint nrpages_out = PAGE_ALIGN(rq->pageofs_out +
+					    rq->outputsize) >> PAGE_SHIFT;
 	unsigned int inputmargin, inlen;
-	u8 *src;
+	u8 *src, *src2;
 	bool copied, support_0padding;
 	int ret;
 
@@ -125,6 +129,7 @@ static int z_erofs_lz4_decompress(struct z_erofs_decompress_req *rq, u8 *out)
 		return -EOPNOTSUPP;
 
 	src = kmap_atomic(*rq->in);
+	src2 = src;
 	inputmargin = 0;
 	support_0padding = false;
 
@@ -148,16 +153,15 @@ static int z_erofs_lz4_decompress(struct z_erofs_decompress_req *rq, u8 *out)
 	if (rq->inplace_io) {
 		const uint oend = (rq->pageofs_out +
 				   rq->outputsize) & ~PAGE_MASK;
-		const uint nr = PAGE_ALIGN(rq->pageofs_out +
-					   rq->outputsize) >> PAGE_SHIFT;
-
 		if (rq->partial_decoding || !support_0padding ||
-		    rq->out[nr - 1] != rq->in[0] ||
+		    rq->out[nrpages_out - 1] != rq->in[0] ||
 		    rq->inputsize - oend <
 		      LZ4_DECOMPRESS_INPLACE_MARGIN(inlen)) {
 			src = generic_copy_inplace_data(rq, src, inputmargin);
 			inputmargin = 0;
 			copied = true;
+		} else {
+			src = obase + ((nrpages_out - 1) << PAGE_SHIFT);
 		}
 	}
 
@@ -187,7 +191,7 @@ static int z_erofs_lz4_decompress(struct z_erofs_decompress_req *rq, u8 *out)
 	if (copied)
 		erofs_put_pcpubuf(src);
 	else
-		kunmap_atomic(src);
+		kunmap_atomic(src2);
 	return ret;
 }
 
@@ -257,7 +261,7 @@ static int z_erofs_decompress_generic(struct z_erofs_decompress_req *rq,
 			return PTR_ERR(dst);
 
 		rq->inplace_io = false;
-		ret = alg->decompress(rq, dst);
+		ret = alg->decompress(rq, dst, NULL);
 		if (!ret)
 			copy_from_pcpubuf(rq->out, dst, rq->pageofs_out,
 					  rq->outputsize);
@@ -291,7 +295,7 @@ static int z_erofs_decompress_generic(struct z_erofs_decompress_req *rq,
 	dst_maptype = 2;
 
 dstmap_out:
-	ret = alg->decompress(rq, dst + rq->pageofs_out);
+	ret = alg->decompress(rq, dst + rq->pageofs_out, dst);
 
 	if (!dst_maptype)
 		kunmap_atomic(dst);
-- 
2.39.3

Re: [PATCH 5.10.y] erofs: fix lz4 inplace decompression

[Greg KH]

On Sat, Feb 24, 2024 at 02:32:48PM +0800, Gao Xiang wrote:
> commit 3c12466b6b7bf1e56f9b32c366a3d83d87afb4de upstream.
> 
> Currently EROFS can map another compressed buffer for inplace
> decompression, that was used to handle the cases that some pages of
> compressed data are actually not in-place I/O.
> 
> However, like most simple LZ77 algorithms, LZ4 expects the compressed
> data is arranged at the end of the decompressed buffer and it
> explicitly uses memmove() to handle overlapping:
>   __________________________________________________________
>  |_ direction of decompression --> ____ |_ compressed data _|
> 
> Although EROFS arranges compressed data like this, it typically maps two
> individual virtual buffers so the relative order is uncertain.
> Previously, it was hardly observed since LZ4 only uses memmove() for
> short overlapped literals and x86/arm64 memmove implementations seem to
> completely cover it up and they don't have this issue.  Juhyung reported
> that EROFS data corruption can be found on a new Intel x86 processor.
> After some analysis, it seems that recent x86 processors with the new
> FSRM feature expose this issue with "rep movsb".
> 
> Let's strictly use the decompressed buffer for lz4 inplace
> decompression for now.  Later, as an useful improvement, we could try
> to tie up these two buffers together in the correct order.
> 
> Reported-and-tested-by: Juhyung Park <qkrwngud825@gmail.com>
> Closes: https://lore.kernel.org/r/CAD14+f2AVKf8Fa2OO1aAUdDNTDsVzzR6ctU_oJSmTyd6zSYR2Q@mail.gmail.com
> Fixes: 0ffd71bcc3a0 ("staging: erofs: introduce LZ4 decompression inplace")
> Fixes: 598162d05080 ("erofs: support decompress big pcluster for lz4 backend")
> Cc: stable <stable@vger.kernel.org> # 5.4+
> Tested-by: Yifan Zhao <zhaoyifan@sjtu.edu.cn>
> Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
> Link: https://lore.kernel.org/r/20231206045534.3920847-1-hsiangkao@linux.alibaba.com
> ---
> Adapt 5.10.y codebase due to non-trivial conflicts out of
> recent new features & cleanups.

Both now queued up, thanks.

greg k-h