Re: [PATCH] btrfs: Fix csum_tree_block to avoid tripping on -Werror=array-bounds

From: David Sterba
Date: Fri May 26 2023 - 10:43:30 EST


On Tue, May 23, 2023 at 09:32:12PM +0200, David Sterba wrote:
> On Tue, May 23, 2023 at 03:33:22PM +0800, Qu Wenruo wrote:
> > On 2023/5/23 15:09, pengfuyuan wrote:
> > Although even with such change, I'm still not sure if it's any better or
> > worse, as most of the calculation can still be bulky.
>
> Yeah I think the calculations would have to be conditional or keeping
> some state. I'd like to keep the structure of the first page and the
> rest.
>
> Possible ways is to add extra condition
>
> for (i = 1; i < num_pages && i < INLINE_EXTENT_BUFFER_PAGES; i++)

The final version is

for (i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++)

ie. 'INLINE_EXTENT_BUFFER_PAGES > 1' can be evaluated at compile time
and result in removing the for loop completely.

Pengfuyuan, can you please do a build test that it does not report the
warning anymore? The diff is:

--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -88,7 +88,6 @@ static void csum_tree_block(struct extent_buffer *buf, u8 *result)
const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize);
SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
char *kaddr;
- int i;

shash->tfm = fs_info->csum_shash;
crypto_shash_init(shash);
@@ -96,7 +95,7 @@ static void csum_tree_block(struct extent_buffer *buf, u8 *result)
crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE,
first_page_part - BTRFS_CSUM_SIZE);

- for (i = 1; i < num_pages; i++) {
+ for (int i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++) {
kaddr = page_address(buf->pages[i]);
crypto_shash_update(shash, kaddr, PAGE_SIZE);
}
---

Thanks.