* [PATCH v1 1/2] crypto: Fix sha1 signed integer comparison compile error
2025-06-14 0:08 [PATCH v1 0/2] crypto: Fix sha1 compile error Yuzhuo Jing
@ 2025-06-14 0:08 ` Yuzhuo Jing
2025-06-14 0:08 ` [PATCH v1 2/2] crypto: Fix sha1 signed pointer " Yuzhuo Jing
2025-06-14 4:53 ` [PATCH v1 0/2] crypto: Fix sha1 " Eric Biggers
2 siblings, 0 replies; 4+ messages in thread
From: Yuzhuo Jing @ 2025-06-14 0:08 UTC (permalink / raw)
To: Herbert Xu, David S . Miller, Ian Rogers, linux-crypto,
linux-kernel
Cc: Yuzhuo Jing
On platforms where -Werror=sign-compare compiler flag is enabled, sha1
code gives errors when for signed to unsigned integer comparisons.
This patch fixes the issue.
Signed-off-by: Yuzhuo Jing <yuzhuo@google.com>
---
include/crypto/sha1_base.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/include/crypto/sha1_base.h b/include/crypto/sha1_base.h
index 0c342ed0d038..3460759d31db 100644
--- a/include/crypto/sha1_base.h
+++ b/include/crypto/sha1_base.h
@@ -73,7 +73,7 @@ static inline int sha1_base_do_update(struct shash_desc *desc,
static inline int sha1_base_do_finalize(struct shash_desc *desc,
sha1_block_fn *block_fn)
{
- const int bit_offset = SHA1_BLOCK_SIZE - sizeof(__be64);
+ const unsigned int bit_offset = SHA1_BLOCK_SIZE - sizeof(__be64);
struct sha1_state *sctx = shash_desc_ctx(desc);
__be64 *bits = (__be64 *)(sctx->buffer + bit_offset);
unsigned int partial = sctx->count % SHA1_BLOCK_SIZE;
@@ -99,7 +99,7 @@ static inline int sha1_base_finish(struct shash_desc *desc, u8 *out)
__be32 *digest = (__be32 *)out;
int i;
- for (i = 0; i < SHA1_DIGEST_SIZE / sizeof(__be32); i++)
+ for (i = 0; i < SHA1_DIGEST_SIZE / (int)sizeof(__be32); i++)
put_unaligned_be32(sctx->state[i], digest++);
memzero_explicit(sctx, sizeof(*sctx));
--
2.50.0.rc1.591.g9c95f17f64-goog
^ permalink raw reply related [flat|nested] 4+ messages in thread* [PATCH v1 2/2] crypto: Fix sha1 signed pointer comparison compile error
2025-06-14 0:08 [PATCH v1 0/2] crypto: Fix sha1 compile error Yuzhuo Jing
2025-06-14 0:08 ` [PATCH v1 1/2] crypto: Fix sha1 signed integer comparison " Yuzhuo Jing
@ 2025-06-14 0:08 ` Yuzhuo Jing
2025-06-14 4:53 ` [PATCH v1 0/2] crypto: Fix sha1 " Eric Biggers
2 siblings, 0 replies; 4+ messages in thread
From: Yuzhuo Jing @ 2025-06-14 0:08 UTC (permalink / raw)
To: Herbert Xu, David S . Miller, Ian Rogers, linux-crypto,
linux-kernel
Cc: Yuzhuo Jing
In include/crypto/sha1_base.h, sha1_block_fn type is defined as
void(sha1_block_fn)(struct sha1_state *sst, u8 const *src, int blocks);
In lib/crypto/sha1.c, the second argument on sha1_transform is defined
as "const char *", which causes type mismatch when calling
sha1_transform from sha1_generic_block_fn in crypto/sha1_generic.c.
We don't break the widely used sha1_block_fn or sha1_transform function
signatures, so this patch converts the pointer sign at usage to fix the
compile error for environments that enable -Werror=pointer-sign.
Signed-off-by: Yuzhuo Jing <yuzhuo@google.com>
---
crypto/sha1_generic.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/crypto/sha1_generic.c b/crypto/sha1_generic.c
index 325b57fe28dc..3a3f9608b989 100644
--- a/crypto/sha1_generic.c
+++ b/crypto/sha1_generic.c
@@ -33,7 +33,7 @@ static void sha1_generic_block_fn(struct sha1_state *sst, u8 const *src,
u32 temp[SHA1_WORKSPACE_WORDS];
while (blocks--) {
- sha1_transform(sst->state, src, temp);
+ sha1_transform(sst->state, (const char *)src, temp);
src += SHA1_BLOCK_SIZE;
}
memzero_explicit(temp, sizeof(temp));
--
2.50.0.rc1.591.g9c95f17f64-goog
^ permalink raw reply related [flat|nested] 4+ messages in thread* Re: [PATCH v1 0/2] crypto: Fix sha1 compile error
2025-06-14 0:08 [PATCH v1 0/2] crypto: Fix sha1 compile error Yuzhuo Jing
2025-06-14 0:08 ` [PATCH v1 1/2] crypto: Fix sha1 signed integer comparison " Yuzhuo Jing
2025-06-14 0:08 ` [PATCH v1 2/2] crypto: Fix sha1 signed pointer " Yuzhuo Jing
@ 2025-06-14 4:53 ` Eric Biggers
2 siblings, 0 replies; 4+ messages in thread
From: Eric Biggers @ 2025-06-14 4:53 UTC (permalink / raw)
To: Yuzhuo Jing
Cc: Herbert Xu, David S . Miller, Ian Rogers, linux-crypto,
linux-kernel
On Fri, Jun 13, 2025 at 05:08:26PM -0700, Yuzhuo Jing wrote:
> This is a followup patch series for an ongoing patch series to reuse
> kernel tree sha1 utils in perf tools and remove libcrypto dependency.
> This mirrors the fixes made in perf back to the kernel tree so we can
> use tools/perf/check-headers.sh to monitor future changes.
> Link: https://lore.kernel.org/lkml/aC9lXhPFcs5fkHWH@x1/t/#u
>
> This series contains two patches: one fixing signed and unsigned integer
> comparisons and another fixing function type mismatches.
>
> Yuzhuo Jing (2):
> crypto: Fix sha1 signed integer comparison compile error
> crypto: Fix sha1 signed pointer comparison compile error
>
> crypto/sha1_generic.c | 2 +-
> include/crypto/sha1_base.h | 4 ++--
> 2 files changed, 3 insertions(+), 3 deletions(-)
I don't like these signedness inconsistencies in the code either, and I'll be
fixing these (among many other issues) when I refactor SHA-1 to have a proper
lib/crypto/ API similar to what I'm currently doing with SHA-2. That being
said, the kernel doesn't have these warnings enabled, and especially in its
current state this code isn't really designed to be copied into a userspace
program.
So I feel that the premise of this patchset, and more importantly also the one
you linked to above for tools/perf/, is a bit misguided.
I've sent an alternative patchset for you to consider:
https://lore.kernel.org/all/20250614044133.660848-1-ebiggers@kernel.org/. It
adds a minimal SHA-1 implementation, including a test, to tools/perf/util/. The
SHA-1 implementation is less than 100 lines anyway.
The effort it would take to "share" the kernel's code here is just not worth it,
IMO. Especially when I have some significant refactoring planned on the kernel
side which would make the tools/perf copy diverge anyway.
- Eric
^ permalink raw reply [flat|nested] 4+ messages in thread