[PATCH 0/2] xfs: check krealloc() return value and simplify code

[PATCH 0/2] xfs: check krealloc() return value and simplify code

[Zhen Lei]

Zhen Lei (2):
  xfs: check the return value of krealloc()
  xfs: remove the extra processing of zero size in xfs_idata_realloc()

 fs/xfs/libxfs/xfs_inode_fork.c | 26 ++++++++++++++++----------
 1 file changed, 16 insertions(+), 10 deletions(-)

-- 
2.26.0.106.g9fadedd

[PATCH 1/2] xfs: check the return value of krealloc()

[Zhen Lei]

krealloc() may fail to expand the memory space. Add sanity checks to it,
and WARN() if that really happened.

Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
---
 fs/xfs/libxfs/xfs_inode_fork.c | 19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_inode_fork.c b/fs/xfs/libxfs/xfs_inode_fork.c
index 7575de5cecb1..4e457aea8493 100644
--- a/fs/xfs/libxfs/xfs_inode_fork.c
+++ b/fs/xfs/libxfs/xfs_inode_fork.c
@@ -366,6 +366,8 @@ xfs_iroot_realloc(
 
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	if (rec_diff > 0) {
+		struct xfs_btree_block *if_broot;
+
 		/*
 		 * If there wasn't any memory allocated before, just
 		 * allocate it now and get out.
@@ -386,8 +388,13 @@ xfs_iroot_realloc(
 		cur_max = xfs_bmbt_maxrecs(mp, ifp->if_broot_bytes, 0);
 		new_max = cur_max + rec_diff;
 		new_size = XFS_BMAP_BROOT_SPACE_CALC(mp, new_max);
-		ifp->if_broot = krealloc(ifp->if_broot, new_size,
-					 GFP_NOFS | __GFP_NOFAIL);
+		if_broot = krealloc(ifp->if_broot, new_size,
+				    GFP_NOFS | __GFP_NOFAIL);
+		if (!if_broot) {
+			WARN(1, "if_broot realloc failed\n");
+			return;
+		}
+		ifp->if_broot = if_broot;
 		op = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, ifp->if_broot, 1,
 						     ifp->if_broot_bytes);
 		np = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, ifp->if_broot, 1,
@@ -477,6 +484,7 @@ xfs_idata_realloc(
 {
 	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
 	int64_t			new_size = ifp->if_bytes + byte_diff;
+	char *if_data;
 
 	ASSERT(new_size >= 0);
 	ASSERT(new_size <= XFS_IFORK_SIZE(ip, whichfork));
@@ -496,8 +504,13 @@ xfs_idata_realloc(
 	 * in size so that it can be logged and stay on word boundaries.
 	 * We enforce that here.
 	 */
-	ifp->if_u1.if_data = krealloc(ifp->if_u1.if_data, roundup(new_size, 4),
+	if_data = krealloc(ifp->if_u1.if_data, roundup(new_size, 4),
 				      GFP_NOFS | __GFP_NOFAIL);
+	if (!if_data) {
+		WARN(1, "if_data realloc failed\n");
+		return;
+	}
+	ifp->if_u1.if_data = if_data;
 	ifp->if_bytes = new_size;
 }
 
-- 
2.26.0.106.g9fadedd

Re: [PATCH 1/2] xfs: check the return value of krealloc()

[Christoph Hellwig]

On Tue, Nov 24, 2020 at 06:45:30PM +0800, Zhen Lei wrote:
> krealloc() may fail to expand the memory space. Add sanity checks to it,
> and WARN() if that really happened.

What part of the __GFP_NOFAIL semantics isn't clear enough?

Re: [PATCH 1/2] xfs: check the return value of krealloc()

[Leizhen (ThunderTown)]

On 2020/11/24 19:51, Christoph Hellwig wrote:
> On Tue, Nov 24, 2020 at 06:45:30PM +0800, Zhen Lei wrote:
>> krealloc() may fail to expand the memory space. Add sanity checks to it,
>> and WARN() if that really happened.
> 
> What part of the __GFP_NOFAIL semantics isn't clear enough?

Oh, sorry, I didn't notice __GFP_NOFAIL flag.

> 
> 

[PATCH 2/2] xfs: remove the extra processing of zero size in xfs_idata_realloc()

[Zhen Lei]

krealloc() does the free operation when the parameter new_size is 0, with
ZERO_SIZE_PTR returned. Because all other places use NULL to check whether
if_data is available or not, so covert it from ZERO_SIZE_PTR to NULL.

Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
---
 fs/xfs/libxfs/xfs_inode_fork.c | 9 +--------
 1 file changed, 1 insertion(+), 8 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_inode_fork.c b/fs/xfs/libxfs/xfs_inode_fork.c
index 4e457aea8493..518af4088e79 100644
--- a/fs/xfs/libxfs/xfs_inode_fork.c
+++ b/fs/xfs/libxfs/xfs_inode_fork.c
@@ -492,13 +492,6 @@ xfs_idata_realloc(
 	if (byte_diff == 0)
 		return;
 
-	if (new_size == 0) {
-		kmem_free(ifp->if_u1.if_data);
-		ifp->if_u1.if_data = NULL;
-		ifp->if_bytes = 0;
-		return;
-	}
-
 	/*
 	 * For inline data, the underlying buffer must be a multiple of 4 bytes
 	 * in size so that it can be logged and stay on word boundaries.
@@ -510,7 +503,7 @@ xfs_idata_realloc(
 		WARN(1, "if_data realloc failed\n");
 		return;
 	}
-	ifp->if_u1.if_data = if_data;
+	ifp->if_u1.if_data = ZERO_OR_NULL_PTR(if_data) ? NULL : if_data;
 	ifp->if_bytes = new_size;
 }
 
-- 
2.26.0.106.g9fadedd

Re: [PATCH 2/2] xfs: remove the extra processing of zero size in xfs_idata_realloc()

[Christoph Hellwig]

On Tue, Nov 24, 2020 at 06:45:31PM +0800, Zhen Lei wrote:
> krealloc() does the free operation when the parameter new_size is 0, with
> ZERO_SIZE_PTR returned. Because all other places use NULL to check whether
> if_data is available or not, so covert it from ZERO_SIZE_PTR to NULL.

This new code looks much harder to read than the version it replaced.

Re: [PATCH 2/2] xfs: remove the extra processing of zero size in xfs_idata_realloc()

[Leizhen (ThunderTown)]

On 2020/11/24 19:52, Christoph Hellwig wrote:
> On Tue, Nov 24, 2020 at 06:45:31PM +0800, Zhen Lei wrote:
>> krealloc() does the free operation when the parameter new_size is 0, with
>> ZERO_SIZE_PTR returned. Because all other places use NULL to check whether
>> if_data is available or not, so covert it from ZERO_SIZE_PTR to NULL.
> 
> This new code looks much harder to read than the version it replaced.

OK

> 
>