[PATCH] staging: lustre: fid: avoid false-positive uninitialized variable warning

[PATCH] staging: lustre: fid: avoid false-positive uninitialized variable warning

[Arnd Bergmann]

One of Neil's recent cleanups apparently has led the code to get
to a state where gcc tracks the 'seqnr' variable just enough to
see that it is sometimes initialized in seq_client_alloc_seq(),
but not enough that it can prove this initialization to be reliable
before the use of that variable:

drivers/staging/lustre/lustre/fid/fid_request.c: In function 'seq_client_alloc_fid':
drivers/staging/lustre/lustre/fid/fid_request.c:245:22: error: 'seqnr' may be used uninitialized in this function [-Werror=maybe-uninitialized]

The code seems to be otherwise correct, and I could not come
up with a good way to simplify it further, so this adds a fake
initialization to shut up that warning.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: NeilBrown <neilb@suse.com>
---
 drivers/staging/lustre/lustre/fid/fid_request.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/staging/lustre/lustre/fid/fid_request.c b/drivers/staging/lustre/lustre/fid/fid_request.c
index fa23423eb8b3..030680f37c79 100644
--- a/drivers/staging/lustre/lustre/fid/fid_request.c
+++ b/drivers/staging/lustre/lustre/fid/fid_request.c
@@ -174,6 +174,7 @@ static int seq_client_alloc_seq(const struct lu_env *env,
 		if (rc) {
 			CERROR("%s: Can't allocate new meta-sequence, rc %d\n",
 			       seq->lcs_name, rc);
+			*seqnr = U64_MAX;
 			return rc;
 		}
 		CDEBUG(D_INFO, "%s: New range - " DRANGE "\n",
-- 
2.9.0

_______________________________________________
devel mailing list
devel@linuxdriverproject.org
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel

Re: [PATCH] staging: lustre: fid: avoid false-positive uninitialized variable warning

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 1838 bytes --]

On Tue, Mar 13 2018, Arnd Bergmann wrote:

> One of Neil's recent cleanups apparently has led the code to get
> to a state where gcc tracks the 'seqnr' variable just enough to
> see that it is sometimes initialized in seq_client_alloc_seq(),
> but not enough that it can prove this initialization to be reliable
> before the use of that variable:
>
> drivers/staging/lustre/lustre/fid/fid_request.c: In function 'seq_client_alloc_fid':
> drivers/staging/lustre/lustre/fid/fid_request.c:245:22: error: 'seqnr' may be used uninitialized in this function [-Werror=maybe-uninitialized]
>
> The code seems to be otherwise correct, and I could not come
> up with a good way to simplify it further, so this adds a fake
> initialization to shut up that warning.

We could possibly have seq_client_alloc_seq return the seqnr,
or U64_MAX+rc.

Then if (seqnr >= U64MAX-MAX_ERRNO) there is an error.

I'm not sure that is actually an improvement though, and I'm happy with
your approach

 Reviewed-by: NeilBrown <neilb@suse.com>

Thanks,
NeilBrown

>
> Cc: NeilBrown <neilb@suse.com>
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> ---
>  drivers/staging/lustre/lustre/fid/fid_request.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/staging/lustre/lustre/fid/fid_request.c b/drivers/staging/lustre/lustre/fid/fid_request.c
> index fa23423eb8b3..030680f37c79 100644
> --- a/drivers/staging/lustre/lustre/fid/fid_request.c
> +++ b/drivers/staging/lustre/lustre/fid/fid_request.c
> @@ -174,6 +174,7 @@ static int seq_client_alloc_seq(const struct lu_env *env,
>  		if (rc) {
>  			CERROR("%s: Can't allocate new meta-sequence, rc %d\n",
>  			       seq->lcs_name, rc);
> +			*seqnr = U64_MAX;
>  			return rc;
>  		}
>  		CDEBUG(D_INFO, "%s: New range - " DRANGE "\n",
> -- 
> 2.9.0

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]