Crypto Update for 2.6.26

Hi Linus:

Here is the crypto update for 2.6.26:

We have just one new algorithm this time with CTS (Cipher Text Stealing). Apart from that it's mostly clean-ups and code reorganisation. Most notbaliy tcrypt is now much smaller in size thanks to Sebastian Siewior's work.

Please pull from

git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

or

master.kernel.org:/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

Alexey Dobriyan (1):
[CRYPTO] api: Switch to proc_create()

David Brownell (1):
[HWRNG] omap: Minor updates

Jan Engelhardt (1):
[CRYPTO] api: Constify function pointer tables

Jan Glauber (4):
[CRYPTO] s390: Generic sha_update and sha_final [CRYPTO] sha512: Hardware acceleration for s390 [CRYPTO] sha384: Hardware acceleration for s390 [CRYPTO] sha512: Rename sha512 to sha512_generic

Kamalesh Babulal (1):
[CRYPTO] all: Clean up init()/fini()

Kevin Coffman (1):
[CRYPTO] cts: Add CTS mode required for Kerberos AES support

Marcin Slusarz (1):
[CRYPTO] lrw: Replace all adds to big endians variables with be*_add_cpu

Sebastian Siewior (10):
[CRYPTO] tcrypt: Group common speed templates [CRYPTO] tcrypt: Shrink speed templates [CRYPTO] aes-x86-32: Remove unused return code [CRYPTO] tcrypt: Change the usage of the test vectors [CRYPTO] tcrypt: Shrink the tcrypt module [CRYPTO] tcrypt: Change the XTEA test vectors [CRYPTO] api: Make the crypto subsystem fully modular [CRYPTO] aes: Export generic setkey
[CRYPTO] padlock-aes: Use generic setkey function [CRYPTO] kconfig: Ordering cleanup

 arch/s390/crypto/Makefile         |    5 
 arch/s390/crypto/crypt_s390.h     |    2 
 arch/s390/crypto/sha.h            |   35 
 arch/s390/crypto/sha1_s390.c      |   91 
 arch/s390/crypto/sha256_s390.c    |   90 
 arch/s390/crypto/sha512_s390.c    |  114 
 arch/s390/crypto/sha_common.c     |   97 
 arch/x86/crypto/aes-i586-asm_32.S |    3 
 crypto/Kconfig                    |  638 +-
 crypto/Makefile                   |    6 
 crypto/aes_generic.c              |   56 
 crypto/anubis.c                   |    8 
 crypto/api.c                      |    3 
 crypto/blowfish.c                 |    8 
 crypto/cast5.c                    |    8 
 crypto/cast6.c                    |    8 
 crypto/crc32c.c                   |    8 
 crypto/crypto_null.c              |    8 
 crypto/cts.c                      |  347 +
 crypto/deflate.c                  |    8 
 crypto/des_generic.c              |    8 
 crypto/fcrypt.c                   |    8 
 crypto/khazad.c                   |    8 
 crypto/lrw.c                      |    5 
 crypto/lzo.c                      |    8 
 crypto/md4.c                      |    8 
 crypto/md5.c                      |    8 
 crypto/proc.c                     |    8 
 crypto/salsa20_generic.c          |    8 
 crypto/serpent.c                  |    8 
 crypto/sha1_generic.c             |    8 
 crypto/sha256_generic.c           |    8 
 crypto/sha512.c                   |  305 
 crypto/sha512_generic.c           |  306 
 crypto/tcrypt.c                   |  529 -
 crypto/tcrypt.h                   |11923 +++++++++++++++++++-------------------
 crypto/tea.c                      |    8 
 crypto/tgr192.c                   |    8 
 crypto/twofish.c                  |    8 
 crypto/wp512.c                    |    8 
 drivers/char/hw_random/omap-rng.c |   26 
 drivers/crypto/Kconfig            |   14 
 drivers/crypto/padlock-aes.c      |  320 -
 include/crypto/aes.h              |    8 
 include/linux/crypto.h            |    7 
 45 files changed, 7710 insertions(+), 7396 deletions(-)

Thanks,

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Hi Linus:

Here is the crypto update for 2.6.27:

The biggest change is the introduction of the new hash interface that is capable of supporting asynchronous implementations as well as reentrancy. There is quite a bit of driver work too in terms of new hardware support and a lot of hifn fixes. On the algorithms side the RIPEMD family (of hash functions) makes a debut. There is also a new CPRNG which has been added for the use of IPsec. And last but not least crc32c has been converted over to the new hash interface in preparation for an Intel-specific implementation.

Please pull from

git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

or

master.kernel.org:/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

Adrian-Ken Rueegsegger (11):
[CRYPTO] ripemd: Add support for RIPEMD hash algorithms [CRYPTO] tcrypt: Add test vectors for RIPEMD-128 and RIPEMD-160 [CRYPTO] ripemd: Add Kconfig entries for RIPEMD hash algorithms [CRYPTO] ripemd: Put all common RIPEMD values in header file [CRYPTO] ripemd: Add support for RIPEMD-256 and RIPEMD-320 [CRYPTO] tcrypt: Add test vectors for RIPEMD-256 and RIPEMD-320 [CRYPTO] ripemd: Add Kconfig entries for extended RIPEMD hash algorithms [CRYPTO] rmd128: Fix endian problems
[CRYPTO] rmd160: Fix endian issues
[CRYPTO] rmd256: Fix endian issues
[CRYPTO] rmd320: Fix endian issues

Christian Hohnstaedt (1):
crypto: ixp4xx - Hardware crypto support for IXP4xx CPUs

Harvey Harrison (3):
crypto: camellia - Use kernel-provided bitops, unaligned access helpers
crypto: rmd128 - sparse annotations
crypto: rmd - sparse annotations

Herbert Xu (10):
[CRYPTO] cryptd: Fix EINPROGRESS notification context [CRYPTO] tcrpyt: Get rid of change log in source [CRYPTO] rmd: Use pointer form of endian swapping operations
crypto: tcrpyt - Remove unnecessary kmap/kunmap calls
crypto: talitos - Use proper form for algorithm driver names
crypto: hash - Fixed digest size check
crypto: hash - Removed vestigial ahash fields
crypto: hash - Added scatter list walking helper
crypto: crc32c - Add ahash implementation
crypto: hash - Move ahash functions into crypto/hash.h

Huang Weiyi (1):
[CRYPTO] hifn: Remove duplicated include

Imre Kaloz (1):
crypto: ixp4xx - Select CRYPTO_AUTHENC

Jan Glauber (1):
crypto: s390 - Respect STFL bit

Jeremy Katz (1):
crypto: padlock - Make module loading quieter when hardware isn't available

Kim Phillips (1):
crypto: talitos - Freescale integrated security engine (SEC) driver

Lee Nipper (2):
crypto: talitos - Add support for 3des
crypto: talitos - Add support for sha256 and md5 variants

Loc Ho (3):
[CRYPTO] hash: Add asynchronous hash support [CRYPTO] cryptd: Add asynchronous hash support [CRYPTO] tcrypt: Use asynchronous hash interface

Neil Horman (2):
[CRYPTO] tcrypt: Add self test for des3_ebe cipher operating in cbc mode
crypto: prng - Deterministic CPRNG

Patrick McHardy (12):
[HIFN]: Endianess fixes
[HIFN]: Remove printk_ratelimit() for debugging printk [HIFN]: Indicate asynchronous processing to crypto API [HIFN]: Handle ablkcipher_walk errors
[HIFN]: Fix data alignment checks
[HIFN]: Properly handle requests for less than the full scatterlist [HIFN]: Use unique driver names for different algos [HIFN]: Properly initialize ivsize for CBC modes
[HIFN]: Fix max queue length value
[HIFN]: Move command descriptor setup to seperate function [HIFN]: Have HW invalidate src and dest descriptors after processing [CRYPTO] tcrypt: Catch cipher destination memory corruption

Robert P. J. Day (1):
[CRYPTO] hifn: Simplify code using ARRAY_SIZE() macro

 arch/s390/crypto/crypt_s390.h  |    4 
 crypto/Kconfig                 |   63 +
 crypto/Makefile                |    7 
 crypto/ahash.c                 |  194 ++++
 crypto/api.c                   |    8 
 crypto/camellia.c              |   84 --
 crypto/crc32c.c                |  128 +++
 crypto/cryptd.c                |  253 ++++++
 crypto/digest.c                |   83 ++
 crypto/hash.c                  |  102 ++
 crypto/hmac.c                  |   16 
 crypto/internal.h              |    1 
 crypto/prng.c                  |  410 ++++++++++
 crypto/prng.h                  |   27 
 crypto/ripemd.h                |   43 +
 crypto/rmd128.c                |  325 ++++++++
 crypto/rmd160.c                |  369 +++++++++
 crypto/rmd256.c                |  344 ++++++++
 crypto/rmd320.c                |  393 ++++++++++
 crypto/tcrypt.c                |  188 +++-
 crypto/tcrypt.h                |  526 +++++++++++++
 drivers/crypto/Kconfig         |   26 
 drivers/crypto/Makefile        |    2 
 drivers/crypto/hifn_795x.c     |  367 ++++-----
 drivers/crypto/ixp4xx_crypto.c | 1506 ++++++++++++++++++++++++++++++++++++++
 drivers/crypto/padlock-aes.c   |    4 
 drivers/crypto/padlock-sha.c   |    4 
 drivers/crypto/talitos.c       | 1597 +++++++++++++++++++++++++++++++++++++++++
 drivers/crypto/talitos.h       |  199 +++++
 include/crypto/hash.h          |  154 +++
 include/crypto/internal/hash.h |   78 ++
 include/linux/crypto.h         |   48 +
 32 files changed, 7227 insertions(+), 326 deletions(-)

Thanks,

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Hi Linus:

Here is the crypto update for 2.6.27:

The biggest change is the introduction of the new hash interface that is capable of supporting asynchronous implementations as well as reentrancy. There is quite a bit of driver work too in terms of new hardware support and a lot of hifn fixes. On the algorithms side the RIPEMD family (of hash functions) makes a debut. There is also a new CPRNG which has been added for the use of IPsec. And last but not least crc32c has been converted over to the new hash interface in preparation for an Intel-specific implementation.

Please pull from

git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

or

master.kernel.org:/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

Adrian-Ken Rueegsegger (11):
[CRYPTO] ripemd: Add support for RIPEMD hash algorithms [CRYPTO] tcrypt: Add test vectors for RIPEMD-128 and RIPEMD-160 [CRYPTO] ripemd: Add Kconfig entries for RIPEMD hash algorithms [CRYPTO] ripemd: Put all common RIPEMD values in header file [CRYPTO] ripemd: Add support for RIPEMD-256 and RIPEMD-320 [CRYPTO] tcrypt: Add test vectors for RIPEMD-256 and RIPEMD-320 [CRYPTO] ripemd: Add Kconfig entries for extended RIPEMD hash algorithms [CRYPTO] rmd128: Fix endian problems
[CRYPTO] rmd160: Fix endian issues
[CRYPTO] rmd256: Fix endian issues
[CRYPTO] rmd320: Fix endian issues

Christian Hohnstaedt (1):
crypto: ixp4xx - Hardware crypto support for IXP4xx CPUs

Harvey Harrison (3):
crypto: camellia - Use kernel-provided bitops, unaligned access helpers
crypto: rmd128 - sparse annotations
crypto: rmd - sparse annotations

Herbert Xu (10):
[CRYPTO] cryptd: Fix EINPROGRESS notification context [CRYPTO] tcrpyt: Get rid of change log in source [CRYPTO] rmd: Use pointer form of endian swapping operations
crypto: tcrpyt - Remove unnecessary kmap/kunmap calls
crypto: talitos - Use proper form for algorithm driver names
crypto: hash - Fixed digest size check
crypto: hash - Removed vestigial ahash fields
crypto: hash - Added scatter list walking helper
crypto: crc32c - Add ahash implementation
crypto: hash - Move ahash functions into crypto/hash.h

Huang Weiyi (1):
[CRYPTO] hifn: Remove duplicated include

Imre Kaloz (1):
crypto: ixp4xx - Select CRYPTO_AUTHENC

Jan Glauber (1):
crypto: s390 - Respect STFL bit

Jeremy Katz (1):
crypto: padlock - Make module loading quieter when hardware isn't available

Kim Phillips (1):
crypto: talitos - Freescale integrated security engine (SEC) driver

Lee Nipper (2):
crypto: talitos - Add support for 3des
crypto: talitos - Add support for sha256 and md5 variants

Loc Ho (3):
[CRYPTO] hash: Add asynchronous hash support [CRYPTO] cryptd: Add asynchronous hash support [CRYPTO] tcrypt: Use asynchronous hash interface

Neil Horman (2):
[CRYPTO] tcrypt: Add self test for des3_ebe cipher operating in cbc mode
crypto: prng - Deterministic CPRNG

Patrick McHardy (12):
[HIFN]: Endianess fixes
[HIFN]: Remove printk_ratelimit() for debugging printk [HIFN]: Indicate asynchronous processing to crypto API [HIFN]: Handle ablkcipher_walk errors
[HIFN]: Fix data alignment checks
[HIFN]: Properly handle requests for less than the full scatterlist [HIFN]: Use unique driver names for different algos [HIFN]: Properly initialize ivsize for CBC modes
[HIFN]: Fix max queue length value
[HIFN]: Move command descriptor setup to seperate function [HIFN]: Have HW invalidate src and dest descriptors after processing [CRYPTO] tcrypt: Catch cipher destination memory corruption

Robert P. J. Day (1):
[CRYPTO] hifn: Simplify code using ARRAY_SIZE() macro

 arch/s390/crypto/crypt_s390.h  |    4 
 crypto/Kconfig                 |   63 +
 crypto/Makefile                |    7 
 crypto/ahash.c                 |  194 ++++
 crypto/api.c                   |    8 
 crypto/camellia.c              |   84 --
 crypto/crc32c.c                |  128 +++
 crypto/cryptd.c                |  253 ++++++
 crypto/digest.c                |   83 ++
 crypto/hash.c                  |  102 ++
 crypto/hmac.c                  |   16 
 crypto/internal.h              |    1 
 crypto/prng.c                  |  410 ++++++++++
 crypto/prng.h                  |   27 
 crypto/ripemd.h                |   43 +
 crypto/rmd128.c                |  325 ++++++++
 crypto/rmd160.c                |  369 +++++++++
 crypto/rmd256.c                |  344 ++++++++
 crypto/rmd320.c                |  393 ++++++++++
 crypto/tcrypt.c                |  188 +++-
 crypto/tcrypt.h                |  526 +++++++++++++
 drivers/crypto/Kconfig         |   26 
 drivers/crypto/Makefile        |    2 
 drivers/crypto/hifn_795x.c     |  367 ++++-----
 drivers/crypto/ixp4xx_crypto.c | 1506 ++++++++++++++++++++++++++++++++++++++
 drivers/crypto/padlock-aes.c   |    4 
 drivers/crypto/padlock-sha.c   |    4 
 drivers/crypto/talitos.c       | 1597 +++++++++++++++++++++++++++++++++++++++++
 drivers/crypto/talitos.h       |  199 +++++
 include/crypto/hash.h          |  154 +++
 include/crypto/internal/hash.h |   78 ++
 include/linux/crypto.h         |   48 +
 32 files changed, 7227 insertions(+), 326 deletions(-)

Thanks,

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Hi Linus:

Here is the crypto update for 2.6.28:

The highlight of this batch is the new run-time self-test framework for algorithms. This reorganises the same tests that we've always had such that they're run at algorithm registration time so that only algorithms passing the tests of their type are made available to users. Hopefully that means people won't be sending me any more algorithm updates that fail these tests :)

Apart from that a new algorithm type for random number generators. The only two algorithms in that type are the existing kernel RNG plus a PRNG based on AES. This is used by IPsec for IV generation.

We also have a new implementation of CRC32c that uses the Intel instruction for CRC32c.

Adrian Bunk (1):
crypto: Kconfig - Replace leading spaces with tabs

Austin Zhang (1):
crypto: crc32c - Use Intel CRC32 instruction

Herbert Xu (10):
crypto: api - Export crypto_alg_lookup instead of __crypto_alg_lookup
crypto: api - Display larval objects properly
crypto: tcrypt - Avoid using contiguous pages
crypto: tcrypt - Abort and only log if there is an error
crypto: tcrypt - Add alg_test interface
crypto: cryptomgr - Add test infrastructure
crypto: api - Use test infrastructure
crypto: cryptomgr - Test ciphers using ECB
crypto: skcipher - Move IV generators into their own modules
crypto: skcipher - Use RNG interface instead of get_random_bytes

Neil Horman (2):
crypto: api - Add fips_enable flag
crypto: rng - RNG interface and implementation

 arch/x86/crypto/Makefile           |    2 
 arch/x86/crypto/crc32c-intel.c     |  197 
 crypto/Kconfig                     |  127 
 crypto/Makefile                    |   12 
 crypto/algapi.c                    |  147 
 crypto/algboss.c                   |  300 +
 crypto/ansi_cprng.c                |  417 +
 crypto/api.c                       |   81 
 crypto/blkcipher.c                 |   29 
 crypto/chainiv.c                   |   44 
 crypto/cryptomgr.c                 |  222 
 crypto/eseqiv.c                    |   35 
 crypto/fips.c                      |   27 
 crypto/internal.h                  |   18 
 crypto/krng.c                      |   66 
 crypto/proc.c                      |   57 
 crypto/rng.c                       |  126 
 crypto/seqiv.c                     |   27 
 crypto/tcrypt.c                    | 1347 -----
 crypto/tcrypt.h                    | 8709 ------------------------------------
 crypto/testmgr.c                   | 1868 +++++++
 crypto/testmgr.h                   | 8738 +++++++++++++++++++++++++++++++++++++
 include/crypto/internal/rng.h      |   26 
 include/crypto/internal/skcipher.h |    6 
 include/crypto/rng.h               |   75 
 include/linux/crypto.h             |   35 
 26 files changed, 12464 insertions(+), 10274 deletions(-)

Thanks,

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Umm. "Here" was empty.

Mind adding where it actually is, again?

Yeah, yeah, I can look at my old merges (you didn't think I _remember_ things, did you?), and I can see that it's going to be

git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6

but really, please do always say where things are supposed to come from!

Also, please us

	git diff -M --stat --summary

for the diffstat. That way I would have seen the

 crypto/{cryptomgr.c => algboss.c}  |   92 +-

as a rename in the email, instead of having to look why my numbers don't match the numbers you claimed..

Linus

Herbert, see section 16 of Documentation/SubmittingPatches.

---
~Randy

Sorry, I somehow deleted that bit while adding the changelog :)

Thanks, I'll run that from now on.

Here it is again:

The highlight of this batch is the new run-time self-test framework for algorithms. This reorganises the same tests that we've always had such that they're run at algorithm registration time so that only algorithms passing the tests of their type are made available to users. Hopefully that means people won't be sending me any more algorithm updates that fail these tests :)

Apart from that a new algorithm type for random number generators. The only two algorithms in that type are the existing kernel RNG plus a PRNG based on AES. This is used by IPsec for IV generation.

We also have a new implementation of CRC32c that uses the Intel instruction for CRC32c.

Please pull from

git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

or

master.kernel.org:/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

Adrian Bunk (1):
crypto: Kconfig - Replace leading spaces with tabs

Austin Zhang (1):
crypto: crc32c - Use Intel CRC32 instruction

Herbert Xu (10):
crypto: api - Export crypto_alg_lookup instead of __crypto_alg_lookup
crypto: api - Display larval objects properly
crypto: tcrypt - Avoid using contiguous pages
crypto: tcrypt - Abort and only log if there is an error
crypto: tcrypt - Add alg_test interface
crypto: cryptomgr - Add test infrastructure
crypto: api - Use test infrastructure
crypto: cryptomgr - Test ciphers using ECB
crypto: skcipher - Move IV generators into their own modules
crypto: skcipher - Use RNG interface instead of get_random_bytes

Neil Horman (2):
crypto: api - Add fips_enable flag
crypto: rng - RNG interface and implementation

 arch/x86/crypto/Makefile           |    2 +
 arch/x86/crypto/crc32c-intel.c     |  197 +
 crypto/Kconfig                     |  127 +-
 crypto/Makefile                    |   12 +-
 crypto/algapi.c                    |  147 +-
 crypto/{cryptomgr.c => algboss.c}  |   92 +-
 crypto/ansi_cprng.c                |  417 ++
 crypto/api.c                       |   81 +-
 crypto/blkcipher.c                 |   29 -
 crypto/chainiv.c                   |   44 +-
 crypto/eseqiv.c                    |   35 +-
 crypto/fips.c                      |   27 +
 crypto/internal.h                  |   18 +-
 crypto/krng.c                      |   66 +
 crypto/proc.c                      |   57 +-
 crypto/rng.c                       |  126 +
 crypto/seqiv.c                     |   27 +-
 crypto/tcrypt.c                    | 1347 +-----
 crypto/tcrypt.h                    | 8709 +-----------------------------------
 crypto/testmgr.c                   | 1868 ++++++++
 crypto/{tcrypt.h => testmgr.h}     |   85 +-
 include/crypto/internal/rng.h      |   26 +
 include/crypto/internal/skcipher.h |    6 -
 include/crypto/rng.h               |   75 +
 include/linux/crypto.h             |   35 +
 25 files changed, 3531 insertions(+), 10124 deletions(-)
 create mode 100644 arch/x86/crypto/crc32c-intel.c
 rename crypto/{cryptomgr.c => algboss.c} (69%)
 create mode 100644 crypto/ansi_cprng.c
 create mode 100644 crypto/fips.c
 create mode 100644 crypto/krng.c
 create mode 100644 crypto/rng.c
 create mode 100644 crypto/testmgr.c
 copy crypto/{tcrypt.h => testmgr.h} (99%)
 create mode 100644 include/crypto/internal/rng.h
 create mode 100644 include/crypto/rng.h

Thanks,

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Noticed the bug below while running our -Werror x86 randconfig build tests against tip/master. Please double check my control flow analysis.

Ingo

---------->
From e5c3b90a18b73d593bba2a3f927a79469b6218d7 Mon Sep 17 00:00:00 2001
From: Ingo Molnar <[email protected]>
Date: Tue, 25 Nov 2008 09:47:28 +0100
Subject: [PATCH] crypto: fix error flow of crypto/testmgr.c:test_comp()
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit

this warning:

crypto/testmgr.c: In function ‘test_comp’:
crypto/testmgr.c:829: warning: ‘ret’ may be used uninitialized in this function

triggers because GCC correctly notices that in the ctcount == 0 && dtcount != 0 input condition case this function can return an undefined value, if the second loop fails.

Remove the shadowed 'ret' variable from the second loop that was probably unintended.

---
 crypto/testmgr.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

index b828c6c..0cf0ae3 100644
@@ -853,7 +853,7 @@ static int test_comp(struct crypto_comp *tfm, struct comp_testvec *ctemplate,
 	}
 
 	for (i = 0; i < dtcount; i++) {
-		int ilen, ret, dlen = COMP_BUF_SIZE;
+		int ilen, dlen = COMP_BUF_SIZE;
 
 		memset(result, 0, sizeof (result));
 

--
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Good catch, I'll apply this patch. Thanks Ingo!

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Thanks.

btw, on a related note, there's a couple of networking related warning fixes/annotations i've got queued up.

I've got 37 commits of that nature in the tip/warnings/* topic branches:

d94b43b: fix warning in drivers/net/s2io.c
78f065d: fix warnings in drivers/net/niu.c
9cc0c65: fix warning in drivers/net/niu.c
32fd3b5: fix warning in drivers/net/qla3xxx.c
ba58621: fix warning in net/sunrpc/xprtrdma/svc_rdma_transport.c
b91eb99: fix warning in net/sched/sch_hfsc.c
c390a37: fix warning in net/netfilter/nf_conntrack_ftp.c 8fecea6: fix warning in net/ipv6/ip6_flowlabel.c
e5e5492: fix warning in drivers/net/wan/z85230.c
67230f2: fix warning in net/dsa/mv88e6xxx.c
948158e: fix warning in net/dsa/mv88e6060.c
8574b3b: fix warning in fs/dlm/netlink.c
9b9c8fd: fix warning in drivers/net/sis900.c
924a289: fix warning in drivers/net/wireless/iwlwifi/iwl-3945.c 2267239: fix warning in drivers/net/wireless/b43/main.c
de96506: fix warning in drivers/net/wireless/ath9k/rc.c 742cddc: fix error path in drivers/net/mlx4/en_rx.c 181787d: fix warning in drivers/net/cassini.c
96c7925: fix warning in net/dccp/options.c
193e56a: fix warning in drivers/net/mlx4/mcg.c
bcc35f7: fix warning in net/ax25/sysctl_net_ax25.c 59681c6: fix warning in net/sunrpc/xprtrdma/verbs.c
ad8d03b: fix warning in net/bluetooth/rfcomm/sock.c 320451a: fix warning in drivers/net/wireless/ray_cs.c
decc165: fix warning in net/netfilter/nf_conntrack_proto_sctp.c
b547d51: fix warning in net/netfilter/ipvs/ip_vs_ctl.c 2281d27: fix warning in drivers/net/atlx/atl2.c
b5a7c2b: fix warning in net/netfilter/nf_conntrack_proto_tcp.c 822c18d: fix warning in drivers/net/wireless/iwlwifi/iwl-scan.c 69f38f6: fix warning in drivers/net/depca.c
557073e: fix warning in drivers/net/3c523.c
d138e44: drivers/net/wireless/b43/phy_g.c: type check debug printouts 7e4f788: fix warning in drivers/net/sky2.c
deb797f: fix warning in net/sunrpc/svcauth_unix.c
923a5a9: fix warning in security/selinux/netnode.c
c62cda5: fix warning in net/packet/af_packet.c
d581867: fix warning in net/rfkill/rfkill.c

Note that the reason there's so many annotations is that this a full range of fixes for allyes/allno/allmod and randconfig kernels on x86. That's a lot of build configurations and a lot of warnings to annotate.

Another one of them is a real bugfix:

742cddc: fix error path in drivers/net/mlx4/en_rx.c

these are side products of -tip's -Werror testing, i.e. not merged anywhere in the direction of linux-next or -git (obviously).

They are done mainly so that we can notice the warnings we introduce via any of the -tip trees - but we analyze and look at all other types of upstream kernel warnings as well.

I'll thus keep these commits indefinitely - but if someone from the networking side would like to work with us to pick them up and eventually nurse them upstream, that would be great!

One thing is not really feasible: there's just too many of them for me to spend 10 minutes on each to submit them with a proper Cc:, etc. (such overhead mounts up quickly if you do the numbers)

So i've attached them below as a dump of patches. If someone wants to cherry-pick bits of them, they are also available in tip/auto-warnings-next:

git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip.git auto-warnings-next

Warning: some of them might be bad, so please double check the analysis.

Ingo

-------------->
commit d94b43b196712688c7b018edd69a4ca1650f61e0

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:55:10 2008 +0100

fix warning in drivers/net/s2io.c

this warning:

drivers/net/s2io.c: In function ‘rx_intr_handler’:
drivers/net/s2io.c:7369: warning: ‘lro’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between:

- s2io_club_tcp_session()and 'lro'

Annotate it.

index 6a1375f..a150575 100644
@@ -7366,7 +7366,7 @@ static int rx_osm_handler(struct ring_info *ring_data, struct RxD_t * rxdp)
 	int ring_no = ring_data->ring_no;
 	u16 l3_csum, l4_csum;
 	unsigned long long err = rxdp->Control_1 & RXD_T_CODE;
-	struct lro *lro;
+	struct lro *uninitialized_var(lro);
 	u8 err_mask;
 
 	skb->dev = dev;

commit 78f065dd3b7f8116dafc7eb6bf966e8276580c15

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:58:10 2008 +0100

fix warnings in drivers/net/niu.c

these warnings:

drivers/net/niu.c: In function ‘serdes_init_niu_1g_serdes’:
drivers/net/niu.c:451: warning: ‘sig’ may be used uninitialized in this function drivers/net/niu.c: In function ‘serdes_init_niu_10g_serdes’:
drivers/net/niu.c:550: warning: ‘sig’ may be used uninitialized in this function

triggers because GCC does not recognize that the max_retry loop always initializes 'sig', due to max_retry != 0.

Annotate them.

index afa399c..7e25142 100644
@@ -448,7 +448,7 @@ static int serdes_init_niu_1g_serdes(struct niu *np)
 	struct niu_link_config *lp = &np->link_config;
 	u16 pll_cfg, pll_sts;
 	int max_retry = 100;
-	u64 sig, mask, val;
+	u64 uninitialized_var(sig), mask, val;
 	u32 tx_cfg, rx_cfg;
 	unsigned long i;
 	int err;

@@ -547,7 +547,7 @@ static int serdes_init_niu_10g_serdes(struct niu *np)
 	struct niu_link_config *lp = &np->link_config;
 	u32 tx_cfg, rx_cfg, pll_cfg, pll_sts;
 	int max_retry = 100;
-	u64 sig, mask, val;
+	u64 uninitialized_var(sig), mask, val;
 	unsigned long i;
 	int err;
 

commit 9cc0c65af74e02da2c3d621cb061bd61b64cc24e

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:54:50 2008 +0100

fix warning in drivers/net/niu.c

this warning:

drivers/net/niu.c: In function ‘esr_reset’:
drivers/net/niu.c:741: warning: ‘reset’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between:

- esr_read_reset() and 'reset'

Annotate it.

index 1b6f548..afa399c 100644
@@ -738,7 +738,7 @@ static int esr_write_glue0(struct niu *np, unsigned long chan, u32 val)
 
 static int esr_reset(struct niu *np)
 {
-	u32 reset;
+	u32 uninitialized_var(reset);
 	int err;
 
 	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,

commit 32fd3b535aa43e290c33873124f91fec6a4329f7

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:37:40 2008 +0100

fix warning in drivers/net/qla3xxx.c

this warning:

drivers/net/qla3xxx.c: In function ‘ql3xxx_probe’:
drivers/net/qla3xxx.c:3912: warning: ‘pci_using_dac’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between 'pci_using_dac' and 'err'.

Annotate it.

index 508452c..3ae623e 100644
@@ -3909,7 +3909,7 @@ static int __devinit ql3xxx_probe(struct pci_dev *pdev,
 	struct net_device *ndev = NULL;
 	struct ql3_adapter *qdev = NULL;
 	static int cards_found = 0;
-	int pci_using_dac, err;
+	int uninitialized_var(pci_using_dac), err;
 
 	err = pci_enable_device(pdev);
 	if (err) {

commit ba58621694ef08b59b753bc5042b3db9472c5aff

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:03:21 2008 +0100

fix warning in net/sunrpc/xprtrdma/svc_rdma_transport.c

this warning:

net/sunrpc/xprtrdma/svc_rdma_transport.c: In function ‘svc_rdma_accept’:
net/sunrpc/xprtrdma/svc_rdma_transport.c:830: warning: ‘dma_mr_acc’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) flow connection between need_dma_mr and dma_mr_acc.

Annotate it.

index 6fb493c..636095c 100644
@@ -827,7 +827,7 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	struct rdma_conn_param conn_param;
 	struct ib_qp_init_attr qp_attr;
 	struct ib_device_attr devattr;
-	int dma_mr_acc;
+	int uninitialized_var(dma_mr_acc);
 	int need_dma_mr;
 	int ret;
 	int i;

commit b91eb990a0d839e810bb15048af08a69ab1b6ffe

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:03:12 2008 +0100

fix warning in net/sched/sch_hfsc.c

this warning:

net/sched/sch_hfsc.c: In function ‘hfsc_enqueue’:
net/sched/sch_hfsc.c:1577: warning: ‘err’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between hfsc_classify(), 'cl' and 'err'.

Annotate it.

index c1e77da..b39e3b5 100644
@@ -1574,7 +1574,7 @@ static int
 hfsc_enqueue(struct sk_buff *skb, struct Qdisc *sch)
 {
 	struct hfsc_class *cl;
-	int err;
+	int uninitialized_var(err);
 
 	cl = hfsc_classify(skb, sch, &err);
 	if (cl == NULL) {

commit c390a3784564c1fb2183e4a31ea39a9beb47d034

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:31:28 2008 +0100

fix warning in net/netfilter/nf_conntrack_ftp.c

this warning:

net/netfilter/nf_conntrack_ftp.c: In function ‘help’:
net/netfilter/nf_conntrack_ftp.c:360: warning: ‘matchoff’ may be used uninitialized in this function net/netfilter/nf_conntrack_ftp.c:360: warning: ‘matchlen’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between find_pattern(), 'found', 'matchoff' and 'matchlen'.

Annotate it.

index 4f71071..af7ad91 100644
@@ -357,7 +357,7 @@ static int help(struct sk_buff *skb,
 	int ret;
 	u32 seq;
 	int dir = CTINFO2DIR(ctinfo);
-	unsigned int matchlen, matchoff;
+	unsigned int uninitialized_var(matchlen), uninitialized_var(matchoff);
 	struct nf_ct_ftp_master *ct_ftp_info = &nfct_help(ct)->help.ct_ftp_info;
 	struct nf_conntrack_expect *exp;
 	union nf_inet_addr *daddr;

@@ -592,3 +592,4 @@ static int __init nf_conntrack_ftp_init(void)
 
 module_init(nf_conntrack_ftp_init);
 module_exit(nf_conntrack_ftp_fini);
+

commit 8fecea67aa434dd1653c4a275532e29903caeb88

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:31:20 2008 +0100

fix warning in net/ipv6/ip6_flowlabel.c

this warning:

net/ipv6/ip6_flowlabel.c: In function ‘ipv6_flowlabel_opt’:
net/ipv6/ip6_flowlabel.c:467: warning: ‘err’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between fl_create() and 'err'.

Annotate it.

index 37a4e77..62809ec 100644
@@ -464,7 +464,7 @@ static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
 
 int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
 {
-	int err;
+	int uninitialized_var(err);
 	struct net *net = sock_net(sk);
 	struct ipv6_pinfo *np = inet6_sk(sk);
 	struct in6_flowlabel_req freq;

commit 67230f214ec140d07fcd32535841ee6eb4437364

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:31:12 2008 +0100

fix warning in net/dsa/mv88e6xxx.c

this warning:

net/dsa/mv88e6xxx.c: In function ‘mv88e6xxx_poll_link’:
net/dsa/mv88e6xxx.c:361: warning: ‘port_status’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between 'link' and 'port_status'.

Annotate it.

index aa6c609..4e4d8b5 100644
@@ -358,7 +358,7 @@ void mv88e6xxx_poll_link(struct dsa_switch *ds)
 
 	for (i = 0; i < DSA_MAX_PORTS; i++) {
 		struct net_device *dev;
-		int port_status;
+		int uninitialized_var(port_status);
 		int link;
 		int speed;
 		int duplex;

commit 948158e35984f86e565db303e0f026e536edca85

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:31:04 2008 +0100

fix warning in net/dsa/mv88e6060.c

this warning:

net/dsa/mv88e6060.c: In function ‘mv88e6060_poll_link’:
net/dsa/mv88e6060.c:225: warning: ‘port_status’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between 'link' and 'port_status'.

Annotate it.

index 54068ef..bfb4a64 100644
@@ -222,7 +222,7 @@ static void mv88e6060_poll_link(struct dsa_switch *ds)
 
 	for (i = 0; i < DSA_MAX_PORTS; i++) {
 		struct net_device *dev;
-		int port_status;
+		int uninitialized_var(port_status);
 		int link;
 		int speed;
 		int duplex;

commit 8574b3b4834c740d37d6eb2d6839037779b56da4

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:30:57 2008 +0100

fix warning in fs/dlm/netlink.c

this warning:

fs/dlm/netlink.c: In function ‘dlm_timeout_warn’:
fs/dlm/netlink.c:131: warning: ‘send_skb’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between prepare_data() and send_skb.

Annotate it.

index 18bda83..aa2a577 100644
@@ -127,8 +127,8 @@ static void fill_data(struct dlm_lock_data *data, struct dlm_lkb *lkb)
 
 void dlm_timeout_warn(struct dlm_lkb *lkb)
 {
+	struct sk_buff *uninitialized_var(send_skb);
 	struct dlm_lock_data *data;
-	struct sk_buff *send_skb;
 	size_t size;
 	int rv;
 

commit 9b9c8fde6c9d175bb7a108e3d231626984364451

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:30:15 2008 +0100

fix warning in drivers/net/sis900.c

this warning:

drivers/net/sis900.c: In function ‘sis900_timer’:
drivers/net/sis900.c:1280: warning: ‘speed’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between sis900_read_mode(), 'speed' and 'duplex'.

Annotate it.

index fa3a460..8be433e 100644
@@ -1277,7 +1277,7 @@ static void sis900_timer(unsigned long data)
 	u16 status;
 
 	if (!sis_priv->autong_complete){
-		int speed, duplex = 0;
+		int uninitialized_var(speed), duplex = 0;
 
 		sis900_read_mode(net_dev, &speed, &duplex);
 		if (duplex){

commit 924a289446dadc9c3c9896c87446e0403c746537

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:38:24 2008 +0100

fix warning in drivers/net/wireless/iwlwifi/iwl-3945.c

this warning:

drivers/net/wireless/iwlwifi/iwl-3945.c: In function ‘iwl3945_txpower_set_from_eeprom’:
drivers/net/wireless/iwlwifi/iwl-3945.c:2182: warning: ‘power_idx’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between iwl3945_hw_reg_get_matched_power_index() and power_idx.

Annotate it.

index 7ca5627..e9e0b1a 100644
@@ -2179,7 +2179,7 @@ int iwl3945_txpower_set_from_eeprom(struct iwl3945_priv *priv)
 		/* set tx power value for all OFDM rates */
 		for (rate_index = 0; rate_index < IWL_OFDM_RATES;
 		     rate_index++) {
-			s32 power_idx;
+			s32 uninitialized_var(power_idx);
 			int rc;
 
 			/* use channel group's clip-power table,

commit 22672393d84dc3c74b7cd46eebdb44d43747318b

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:38:08 2008 +0100

fix warning in drivers/net/wireless/b43/main.c

this warning:

drivers/net/wireless/b43/main.c: In function ‘b43_op_config’:
drivers/net/wireless/b43/main.c:3258: warning: ‘gmode’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between the list_for_each_entry() loop, up_dev and gmode.

Annotate it.

index 14c44df..bd70b27 100644
@@ -3255,7 +3255,7 @@ static int b43_switch_band(struct b43_wl *wl, struct ieee80211_channel *chan)
 	struct b43_wldev *down_dev;
 	struct b43_wldev *d;
 	int err;
-	bool gmode;
+	bool uninitialized_var(gmode);
 	int prev_status;
 
 	/* Find a device and PHY which supports the band. */

commit de965062d615afc885bd800976e488bd3c2d1d6f

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:37:58 2008 +0100

fix warning in drivers/net/wireless/ath9k/rc.c

this warning:

drivers/net/wireless/ath9k/rc.c: In function ‘ath_rc_rate_getidx’:
drivers/net/wireless/ath9k/rc.c:1033: warning: ‘nextindex’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between ath_rc_get_nextlowervalid_txrate() and &nextindex.

Annotate it.

index cca2fc5..27b35d3 100644
@@ -1030,7 +1030,7 @@ static u8 ath_rc_rate_getidx(struct ath_softc *sc,
 			     u16 min_rate)
 {
 	u32 j;
-	u8 nextindex;
+	u8 uninitialized_var(nextindex);
 	struct ath_tx_ratectrl *rate_ctrl =
 		(struct ath_tx_ratectrl *)(ath_rc_priv);
 

commit e5e5492e7094ed4bea8daf0302f60fbb21daa4c7

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 14:52:20 2008 +0100

fix warning in drivers/net/wan/z85230.c

this warning:

drivers/net/wan/z85230.c: In function ‘z8530_interrupt’:
drivers/net/wan/z85230.c:713: warning: ‘intr’ may be used uninitialized in this function

is clearly bogus - annotate it.

index 5bf7e01..d7bf53a 100644
@@ -710,7 +710,7 @@ EXPORT_SYMBOL(z8530_nop);
 irqreturn_t z8530_interrupt(int irq, void *dev_id)
 {
 	struct z8530_dev *dev=dev_id;
-	u8 intr;
+	u8 uninitialized_var(intr);
 	static volatile int locker=0;
 	int work=0;
 	struct z8530_irqhandler *irqs;

commit 742cddc749294be3c0fa909e6255c5a338a1ac40

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:37:47 2008 +0100

fix error path in drivers/net/mlx4/en_rx.c

this warning:

drivers/net/mlx4/en_rx.c: In function ‘mlx4_en_activate_rx_rings’:
drivers/net/mlx4/en_rx.c:412: warning: ‘err’ may be used uninitialized in this function

Triggers because 'err' is uninitialized in the following input
conditions: priv->rx_ring_num is zero and mlx4_en_fill_rx_buffers() fails.

But even if ->rx_ring_num is nonzero, 'err' will be zero if mlx4_en_fill_rx_buffers() fails and mlx4_en_activate_rx_rings() returns success - incorrectly.

So it's best to keep the error code uptodate on mlx4_en_fill_rx_buffers() calls as well.

index 6232227..aed7739 100644
@@ -443,7 +443,8 @@ int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv)
 		/* Fill Rx buffers */
 		ring->full = 0;
 	}
-	if (mlx4_en_fill_rx_buffers(priv))
+	err = mlx4_en_fill_rx_buffers(priv);
+	if (err)
 		goto err_buffers;
 
 	for (ring_ind = 0; ring_ind < priv->rx_ring_num; ring_ind++) {

commit 181787d1c30af364ffd15428fe6fab7798c3f3c1

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:37:35 2008 +0100

fix warning in drivers/net/cassini.c

this warning:

drivers/net/cassini.c: In function ‘cas_rx_ringN’:
drivers/net/cassini.c:2350: warning: ‘skb’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between cas_rx_process_pkt() and 'skb'.

Annotate it.

index 86909cf..e083e1c 100644
@@ -2347,7 +2347,7 @@ static int cas_rx_ringN(struct cas *cp, int ring, int budget)
 	drops = 0;
 	while (1) {
 		struct cas_rx_comp *rxc = rxcs + entry;
-		struct sk_buff *skb;
+		struct sk_buff *uninitialized_var(skb);
 		int type, len;
 		u64 words[4];
 		int i, dring;

commit 96c79254264033f1505198d5d362b68d88d1bbaf

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 12:41:30 2008 +0200

fix warning in net/dccp/options.c

this warning:

net/dccp/options.c: In function ‘dccp_parse_options’:
net/dccp/options.c:67: warning: ‘value’ may be used uninitialized in this function

is a bogus GCC warning. The compiler does not recognize the relation between "value" and "mandatory" variables: the code flow can ever reach the "out_invalid_option:" label if 'mandatory' is set to 1, and when 'mandatory' is non-zero, we'll always have 'value' initialized.

Help out the compiler by annotating the variable.

index 0809b63..18dcfb9 100644
@@ -64,7 +64,7 @@ int dccp_parse_options(struct sock *sk, struct dccp_request_sock *dreq,
 					(dh->dccph_doff * 4);
 	struct dccp_options_received *opt_recv = &dp->dccps_options_received;
 	unsigned char opt, len;
-	unsigned char *value;
+	unsigned char *uninitialized_var(value);
 	u32 elapsed_time;
 	__be32 opt_val;
 	int rc;

commit 193e56a10e56505dc7eac9e5632834c72b55613d

Author: Ingo Molnar <[email protected]>
Date:   Mon Aug 18 16:18:34 2008 +0200

fix warning in drivers/net/mlx4/mcg.c

fix warning:

drivers/net/mlx4/mcg.c: In function ‘mlx4_multicast_attach’:
drivers/net/mlx4/mcg.c:217: warning: integer overflow in expression

there was no real danger of overflow here though.

md5:
db8eb55620f886c03854a2abb2ce6c3f mcg.o.before.asm db8eb55620f886c03854a2abb2ce6c3f mcg.o.after.asm

index 592c01a..224ef0f 100644
@@ -215,7 +215,7 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
 
 	if (block_mcast_loopback)
 		mgm->qp[members_count++] = cpu_to_be32((qp->qpn & MGM_QPN_MASK) |
-						       (1 << MGM_BLCK_LB_BIT));
+						       (1U << MGM_BLCK_LB_BIT));
 	else
 		mgm->qp[members_count++] = cpu_to_be32(qp->qpn & MGM_QPN_MASK);
 

commit bcc35f75a3b9b96c53a74176b135304fc7ffe052

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 15:49:15 2008 +0200

fix warning in net/ax25/sysctl_net_ax25.c

fix this warning:

net/ax25/sysctl_net_ax25.c:27: warning: ‘min_ds_timeout’ defined but not used net/ax25/sysctl_net_ax25.c:27: warning: ‘max_ds_timeout’ defined but not used

These are only used in the CONFIG_AX25_DAMA_SLAVE case.

index f288fc4..735ceef 100644
@@ -24,7 +24,9 @@ static int min_idle[1],			max_idle[] = {65535000};
 static int min_n2[] = {1},		max_n2[] = {31};
 static int min_paclen[] = {1},		max_paclen[] = {512};
 static int min_proto[1],		max_proto[] = { AX25_PROTO_MAX };
+#ifdef CONFIG_AX25_DAMA_SLAVE
 static int min_ds_timeout[1],		max_ds_timeout[] = {65535000};
+#endif
 
 static struct ctl_table_header *ax25_table_header;
 

commit 59681c6c30e3f700af8d716f66e41ad18f726b80

Author: Ingo Molnar <[email protected]>
Date:   Mon Oct 20 18:50:11 2008 +0200

fix warning in net/sunrpc/xprtrdma/verbs.c

fix this warning:

net/sunrpc/xprtrdma/verbs.c: In function ‘rpcrdma_conn_upcall’:
net/sunrpc/xprtrdma/verbs.c:279: warning: unused variable ‘addr’

index a5fef5e..900e595 100644
@@ -276,7 +276,9 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 	struct rpcrdma_xprt *xprt = id->context;
 	struct rpcrdma_ia *ia = &xprt->rx_ia;
 	struct rpcrdma_ep *ep = &xprt->rx_ep;
+#ifdef RPC_DEBUG
 	struct sockaddr_in *addr = (struct sockaddr_in *) &ep->rep_remote_addr;
+#endif
 	struct ib_qp_attr attr;
 	struct ib_qp_init_attr iattr;
 	int connstate = 0;

commit ad8d03b315507a954160fcf04dc94cef9f118e77

Author: Ingo Molnar <[email protected]>
Date:   Mon Oct 20 17:34:02 2008 +0200

fix warning in net/bluetooth/rfcomm/sock.c

fix this warning:

net/bluetooth/rfcomm/sock.c: In function ‘rfcomm_sock_ioctl’:
net/bluetooth/rfcomm/sock.c:795: warning: unused variable ‘sk’

perhaps BT_DEBUG() should be improved to do printf format checking instead of the #ifdef, but that looks quite intrusive: each bluetooth .c file undefines the macro.

index 8a972b6..bc0d4a7 100644
@@ -792,7 +792,9 @@ static int rfcomm_sock_getsockopt(struct socket *sock, int level, int optname, c
 
 static int rfcomm_sock_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 {
+#if defined(CONFIG_BT_RFCOMM_TTY) || defined(CONFIG_BT_RFCOMM_DEBUG)
 	struct sock *sk = sock->sk;
+#endif
 	int err;
 
 	BT_DBG("sk %p cmd %x arg %lx", sk, cmd, arg);

commit 320451a3f79098ca800f2e58674074c26efee020

Author: Ingo Molnar <[email protected]>
Date:   Sat Oct 18 21:33:06 2008 +0200

fix warning in drivers/net/wireless/ray_cs.c

fix this warning:

drivers/net/wireless/ray_cs.c:297: warning: ‘rcsid’ defined but not used

'rcsid' is only used in the CONFIG_PROC_FS=y case.

index 1404a57..c23a811 100644
@@ -294,7 +294,9 @@ static char hop_pattern_length[] = { 1,
 	     JAPAN_TEST_HOP_MOD
 };
 
+#ifdef CONFIG_PROC_FS
 static char rcsid[] = "Raylink/WebGear wireless LAN - Corey <Thomas [email protected]>";
+#endif
 
 /*=============================================================================
     ray_attach() creates an "instance" of the driver, allocating

commit decc1656e870c9e5bd94f2357b8496b8da6c2e73

Author: Ingo Molnar <[email protected]>
Date:   Sat Oct 18 19:14:40 2008 +0200

fix warning in net/netfilter/nf_conntrack_proto_sctp.c

fix this warning:

net/netfilter/nf_conntrack_proto_sctp.c: In function ‘sctp_packet’:
net/netfilter/nf_conntrack_proto_sctp.c:376: warning: array subscript is above array bounds

GCC noticed that new_state could in theory be SCTP_CONNTRACK_MAX, which is beyond the limit of the sctp_timeouts[] array.

This situation could happen if for_each_sctp_chunk() exits immediately. That should not normally happen - but it's not obvious to me and it's better to insert an extra check to the end of the loop for this case.

This gets rid of warning as well.

index ae8c260..d31ced4 100644
@@ -373,6 +373,9 @@ static int sctp_packet(struct nf_conn *ct,
 	}
 	write_unlock_bh(&sctp_lock);
 
+	if (new_state == SCTP_CONNTRACK_MAX)
+		goto out;
+
 	nf_ct_refresh_acct(ct, ctinfo, skb, sctp_timeouts[new_state]);
 
 	if (old_state == SCTP_CONNTRACK_COOKIE_ECHOED &&

commit b547d51a26c5ed23571df50c7254439e2f5dbebe

Author: Ingo Molnar <[email protected]>
Date:   Sat Oct 18 19:02:37 2008 +0200

fix warning in net/netfilter/ipvs/ip_vs_ctl.c

Impact: fix potential kernel stack information exposure to user-space.

fix this warning:

net/netfilter/ipvs/ip_vs_ctl.c: In function ‘ip_vs_genl_set_cmd’:
net/netfilter/ipvs/ip_vs_ctl.c:3034: warning: ‘t.udp_timeout’ may be used uninitialized in this function net/netfilter/ipvs/ip_vs_ctl.c:3034: note: ‘t.udp_timeout’ was declared here

GCC seems to be correct here: in the !CONFIG_IP_VS_PROTO_UDP case __ip_vs_get_timeouts keeps the timeout structure uninitialized. But in a couple of places we use and copy this structure to user-space, potentially exposing kernel-stack data to user-space.

While it's a minor issue that only triggers with certain config variations, it's best to always initialize this structure in __ip_vs_get_timeouts(). (GCC will eliminate this memset)

index 0302cf3..7e3a0c7 100644
@@ -2327,6 +2327,7 @@ __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
 static inline void
 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
 {
+	memset(u, 0, sizeof(*u));
 #ifdef CONFIG_IP_VS_PROTO_TCP
 	u->tcp_timeout =
 		ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;

commit 2281d27663a5f17689ec49e029b4962cba19d614

Author: Ingo Molnar <[email protected]>
Date:   Sat Oct 18 16:27:37 2008 +0200

fix warning in drivers/net/atlx/atl2.c

fix this warning:

drivers/net/atlx/atl2.c: In function ‘atl2_request_irq’:
drivers/net/atlx/atl2.c:644: warning: unused variable ‘err’

'err' is unused in the !CONFIG_PCI_MSI case.

Instead of further increasing the #ifdeffery in this function, restructure the code a bit and get rid of the #ifdef. This relies on the fact that pci_enable_msi() will always fail in the !CONFIG_PCI_MSI case.

There should be no change in driver behavior.

index 8571e8c..e02bd7e 100644
@@ -644,7 +644,6 @@ static int atl2_request_irq(struct atl2_adapter *adapter)
 	int flags, err = 0;
 
 	flags = IRQF_SHARED;
-#ifdef CONFIG_PCI_MSI
 	adapter->have_msi = true;
 	err = pci_enable_msi(adapter->pdev);
 	if (err)

@@ -652,7 +651,6 @@ static int atl2_request_irq(struct atl2_adapter *adapter)
 
 	if (adapter->have_msi)
 		flags &= ~IRQF_SHARED;
-#endif
 
 	return request_irq(adapter->pdev->irq, &atl2_intr, flags, netdev->name,
 		netdev);

commit b5a7c2b9e1a1b4d80da5d66b330545fdeafe1dad

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 19:48:51 2008 +0200

fix warning in net/netfilter/nf_conntrack_proto_tcp.c

fix this warning:

net/netfilter/nf_conntrack_proto_tcp.c: In function ‘tcp_in_window’:
net/netfilter/nf_conntrack_proto_tcp.c:491: warning: unused variable ‘net’ net/netfilter/nf_conntrack_proto_tcp.c: In function ‘tcp_packet’:
net/netfilter/nf_conntrack_proto_tcp.c:812: warning: unused variable ‘net’

index 7f2f43c..debdaf7 100644
@@ -129,7 +129,7 @@ extern const struct nla_policy nf_ct_port_nla_policy[];
 	 && net_ratelimit())
 #endif
 #else
-#define LOG_INVALID(net, proto) 0
+static inline int LOG_INVALID(struct net *net, int proto) { return 0; }
 #endif /* CONFIG_SYSCTL */
 
 #endif /*_NF_CONNTRACK_PROTOCOL_H*/

commit 822c18db871bc88b2a60d7a9618ec79a8b5e1ef5

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 18:31:33 2008 +0200

fix warning in drivers/net/wireless/iwlwifi/iwl-scan.c

fix this warning:

drivers/net/wireless/iwlwifi/iwl-scan.c:92: warning: ‘iwl_escape_essid’ defined but not used

Convert the IWL debug macros to proper printf format checking. Besides better type checking, this gets rid of the warning as well.

index e548d67..ddc1fbf 100644
@@ -68,8 +68,10 @@ void iwl_dbgfs_unregister(struct iwl_priv *priv);
 #endif
 
 #else
-#define IWL_DEBUG(level, fmt, args...)
-#define IWL_DEBUG_LIMIT(level, fmt, args...)
+static inline void __attribute__ ((format(printf, 2, 3)))
+IWL_DEBUG(int level, const char *fmt, ...) { }
+static inline void __attribute__ ((format(printf, 2, 3)))
+IWL_DEBUG_LIMIT(int level, const char *fmt, ...) { }
 #endif				/* CONFIG_IWLWIFI_DEBUG */
 
 

index 4eee1b1..f4f0664 100644
@@ -43,7 +43,6 @@
 #include "iwl-io.h"
 #include "iwl-helpers.h"
 
-#ifdef CONFIG_IWLWIFI_DEBUG
 static const char *led_type_str[] = {
 	__stringify(IWL_LED_TRG_TX),
 	__stringify(IWL_LED_TRG_RX),

@@ -51,7 +50,6 @@ static const char *led_type_str[] = {
 	__stringify(IWL_LED_TRG_RADIO),
 	NULL
 };
-#endif /* CONFIG_IWLWIFI_DEBUG */
 
 
 static const struct {

commit 69f38f617c3f832e088be9d35242d929aa623fd4

Author: Ingo Molnar <[email protected]>
Date:   Wed Aug 27 19:06:48 2008 +0200

fix warning in drivers/net/depca.c

fix warning:

drivers/net/depca.c: In function ‘depca_eisa_probe’:
drivers/net/depca.c:1564: warning: ‘mem_start’ may be used uninitialized in this function

this seems to be a real bug - depca_eisa_probe() does not check for failure. Add it, symmetric to depca_isa_probe().

index ace39ec..f00bd64 100644
@@ -1558,6 +1558,7 @@ static int __init depca_isa_probe (struct platform_device *device)
 #ifdef CONFIG_EISA
 static int __init depca_eisa_probe (struct device *device)
 {
+	enum depca_type adapter = unknown;
 	struct eisa_device *edev;
 	struct net_device *dev;
 	struct depca_private *lp;

@@ -1576,7 +1577,11 @@ static int __init depca_eisa_probe (struct device *device)
 	 * the EISA configuration structures (yet... :-), just rely on
 	 * the ISA probing to sort it out... */
 
-	depca_shmem_probe (&mem_start);
+	adapter = depca_shmem_probe (&mem_start);
+	if (adapter == unknown) {
+		status = -ENODEV;
+		goto out_free;
+	}
 
 	dev->base_addr = ioaddr;
 	dev->irq = irq;

commit 557073e3dac7b8479df69c140a5fd2b8f4a9cfad

Author: Ingo Molnar <[email protected]>
Date:   Wed Aug 27 19:05:19 2008 +0200

fix warning in drivers/net/3c523.c

fix warning:

drivers/net/3c523.c:582: warning: ‘cleanup_card’ defined but not used

No code changed:

md5:
ebe4a1b27d3f21b0b12a78c58463b0d7 3c523.o.before.asm ebe4a1b27d3f21b0b12a78c58463b0d7 3c523.o.after.asm

index e2ce41d..7b9d427 100644
@@ -578,13 +578,13 @@ err_out:
 	return retval;
 }
 
+#ifdef MODULE
 static void cleanup_card(struct net_device *dev)
 {
 	mca_set_adapter_procfn(((struct priv *) (dev->priv))->slot, NULL, NULL);
 	release_region(dev->base_addr, ELMC_IO_EXTENT);
 }
-
-#ifndef MODULE
+#else
 struct net_device * __init elmc_probe(int unit)
 {
 	struct net_device *dev = alloc_etherdev(sizeof(struct priv));

commit d138e44e2f8d0f26e04b0386d328d9cc7e33d82e

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 14:30:37 2008 +0200

drivers/net/wireless/b43/phy_g.c: type check debug printouts

this warning:

drivers/net/wireless/b43/phy_g.c: In function ‘b43_gphy_op_recalc_txpower’:
drivers/net/wireless/b43/phy_g.c:3191: warning: unused variable ‘dbm’

is caused because b43dbg() is a macro, hence in the !B43_DEBUG case the compiler does not know that the variables are used.

Convert it to a proper inline instead. This also improves type checking in the !B43_DEBUG case.

index 427b820..ac55b62 100644
@@ -853,7 +853,8 @@ void b43warn(struct b43_wl *wl, const char *fmt, ...)
 void b43dbg(struct b43_wl *wl, const char *fmt, ...)
     __attribute__ ((format(printf, 2, 3)));
 #else /* DEBUG */
-# define b43dbg(wl, fmt...) do { /* nothing */ } while (0)
+static inline void __attribute__ ((format(printf, 2, 3)))
+b43dbg(struct b43_wl *wl, const char *fmt, ...) { }
 #endif /* DEBUG */
 
 /* A WARN_ON variant that vanishes when b43 debugging is disabled.

commit 7e4f7889aca92f2cf95ba31e005fc1368c1cbc5a

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 17:47:04 2008 +0200

fix warning in drivers/net/sky2.c

fix this warning:

drivers/net/sky2.c: In function ‘sky2_mhz’:
drivers/net/sky2.c:2753: warning: control reaches end of non-void function

BUG() can be NOP on !CONFIG_BUG.

index 3813d15..85fdacc 100644
@@ -2748,7 +2748,7 @@ static u32 sky2_mhz(const struct sky2_hw *hw)
 		return 156;
 
 	default:
-		BUG();
+		panic("sky2_mhz: unknown chip id!");
 	}
 }
 

commit deb797f96d920083ba8a0bcfa6f57b19fc593571

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 17:34:27 2008 +0200

fix warning in net/sunrpc/svcauth_unix.c

fix this warning:

net/sunrpc/svcauth_unix.c: In function ‘svcauth_unix_set_client’:
net/sunrpc/svcauth_unix.c:681: warning: ‘sin6’ may be used uninitialized in this function

caused by BUG() being a NOP on !CONFIG_BUG.

index f24800f..200ed93 100644
@@ -692,7 +692,7 @@ svcauth_unix_set_client(struct svc_rqst *rqstp)
 		sin6 = svc_addr_in6(rqstp);
 		break;
 	default:
-		BUG();
+		panic("svcauth_unix_set_client: bad address family!");
 	}
 
 	rqstp->rq_client = NULL;

@@ -873,3 +873,4 @@ struct auth_ops svcauth_unix = {
 	.set_client	= svcauth_unix_set_client,
 };
 
+

commit 923a5a960577dcc5189b6113300a122c0212b292

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 16:55:34 2008 +0200

fix warning in security/selinux/netnode.c

fix this warning:

security/selinux/netnode.c: In function ‘sel_netnode_find’:
security/selinux/netnode.c:131: warning: ‘idx’ may be used uninitialized in this function

BUG() can be a NOP in the !CONFIG_BUG case. Return NULL in that case, instead of running in an undefined way.

index 7100072..a307122 100644
@@ -140,6 +140,7 @@ static struct sel_netnode *sel_netnode_find(const void *addr, u16 family)
 		break;
 	default:
 		BUG();
+		return NULL;
 	}
 
 	list_for_each_entry_rcu(node, &sel_netnode_hash[idx].list, list)

commit c62cda5dfaf7630edefe097a3e23af7732060b08

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 16:26:02 2008 +0200

fix warning in net/packet/af_packet.c

fix this warning:

net/packet/af_packet.c: In function ‘tpacket_rcv’:
net/packet/af_packet.c:604: warning: ‘hdrlen’ may be used uninitialized in this function

This is because in the !CONFIG_BUG case BUG() is a NOP.

There is no other sane way to continue but to panic the system unconditionally, regardless of the CONFIG_BUG setting.

index c718e7e..1478fb8 100644
@@ -709,7 +709,7 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, struct packe
 		hdrlen = sizeof(*h.h2);
 		break;
 	default:
-		BUG();
+		panic("AF_PACKET: bad tp->version");
 	}
 
 	sll = h.raw + TPACKET_ALIGN(hdrlen);

commit d581867031ebe3c6f4e018ea32ed8b9ee5141ead

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 16:15:59 2008 +0200

fix warning in net/rfkill/rfkill.c

fix this warning:

net/rfkill/rfkill.c: In function ‘rfkill_get_type_str’:
net/rfkill/rfkill.c:393: warning: control reaches end of non-void function

in the embedded !CONFIG_BUG case BUG() is a NOP.

Return NULL to solve this (unlikely) scenario. All callers feed this pointer into sprintf %s so NULL should be fine.

index f949a48..eecb9a8 100644
@@ -389,6 +389,7 @@ static const char *rfkill_get_type_str(enum rfkill_type type)
 		return "wwan";
 	default:
 		BUG();
+		return NULL;
 	}
 }
 

--
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

I've picked up this one in my nf-next-2.6.git tree.

This one is already fixed differently in my tree.

I've also picked up this one.

From: Ingo Molnar <[email protected]>
Date: Tue, 25 Nov 2008 18:14:38 +0100

I see these now, but they would get properly logged and tracked at:

http://patchwork.ozlabs.org/project/netdev/

if you would simply CC: netdev instead of sending them into the ether by informing lkml and the crypto list about them.

sure - netdev Cc:-ed, subject line changed - mail repeated below.

This one seems to be a real bug too (subject line nonwithstanding):

69f38f6: fix warning in drivers/net/depca.c

besides the bugfixes, the annotations would be nice to have too - i got the warnings with latest-gcc (4.3.2) and once we reach a baseline of no warnings it will be much easier to go forward in a delta manner. At that points commits that actually cause new warnings will be much fresher - and the warnings will be a lot more interesting in general as well.

Right now we've got a baseline noise where it's mostly GCC that is wrong. Obviously those are the warnings that tend to linger around in the kernel source: people see the warnings and determine that GCC is wrong and keep the source unchanged.

The arch/x86/* files already have a zero warnings baseline (clean -Werror builds), and my experience is that the new warnings that pop up are generally quite healthy and interesting, with a low false positive rate.

Hm?

Ingo

---------------->

I've got 37 commits of that nature in the tip/warnings/* topic branches:

d94b43b: fix warning in drivers/net/s2io.c
78f065d: fix warnings in drivers/net/niu.c
9cc0c65: fix warning in drivers/net/niu.c
32fd3b5: fix warning in drivers/net/qla3xxx.c
ba58621: fix warning in net/sunrpc/xprtrdma/svc_rdma_transport.c
b91eb99: fix warning in net/sched/sch_hfsc.c
c390a37: fix warning in net/netfilter/nf_conntrack_ftp.c 8fecea6: fix warning in net/ipv6/ip6_flowlabel.c
e5e5492: fix warning in drivers/net/wan/z85230.c
67230f2: fix warning in net/dsa/mv88e6xxx.c
948158e: fix warning in net/dsa/mv88e6060.c
8574b3b: fix warning in fs/dlm/netlink.c
9b9c8fd: fix warning in drivers/net/sis900.c
924a289: fix warning in drivers/net/wireless/iwlwifi/iwl-3945.c 2267239: fix warning in drivers/net/wireless/b43/main.c
de96506: fix warning in drivers/net/wireless/ath9k/rc.c 742cddc: fix error path in drivers/net/mlx4/en_rx.c 181787d: fix warning in drivers/net/cassini.c
96c7925: fix warning in net/dccp/options.c
193e56a: fix warning in drivers/net/mlx4/mcg.c
bcc35f7: fix warning in net/ax25/sysctl_net_ax25.c 59681c6: fix warning in net/sunrpc/xprtrdma/verbs.c
ad8d03b: fix warning in net/bluetooth/rfcomm/sock.c 320451a: fix warning in drivers/net/wireless/ray_cs.c
decc165: fix warning in net/netfilter/nf_conntrack_proto_sctp.c
b547d51: fix warning in net/netfilter/ipvs/ip_vs_ctl.c 2281d27: fix warning in drivers/net/atlx/atl2.c
b5a7c2b: fix warning in net/netfilter/nf_conntrack_proto_tcp.c 822c18d: fix warning in drivers/net/wireless/iwlwifi/iwl-scan.c 69f38f6: fix warning in drivers/net/depca.c
557073e: fix warning in drivers/net/3c523.c
d138e44: drivers/net/wireless/b43/phy_g.c: type check debug printouts 7e4f788: fix warning in drivers/net/sky2.c
deb797f: fix warning in net/sunrpc/svcauth_unix.c
923a5a9: fix warning in security/selinux/netnode.c
c62cda5: fix warning in net/packet/af_packet.c
d581867: fix warning in net/rfkill/rfkill.c

Note that the reason there's so many annotations is that this a full range of fixes for allyes/allno/allmod and randconfig kernels on x86. That's a lot of build configurations and a lot of warnings to annotate.

Another one of them is a real bugfix:

742cddc: fix error path in drivers/net/mlx4/en_rx.c

these are side products of -tip's -Werror testing, i.e. not merged anywhere in the direction of linux-next or -git (obviously).

They are done mainly so that we can notice the warnings we introduce via any of the -tip trees - but we analyze and look at all other types of upstream kernel warnings as well.

I'll thus keep these commits indefinitely - but if someone from the networking side would like to work with us to pick them up and eventually nurse them upstream, that would be great!

One thing is not really feasible: there's just too many of them for me to spend 10 minutes on each to submit them with a proper Cc:, etc. (such overhead mounts up quickly if you do the numbers)

So i've attached them below as a dump of patches. If someone wants to cherry-pick bits of them, they are also available in tip/auto-warnings-next:

git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip.git auto-warnings-next

Warning: some of them might be bad, so please double check the analysis.

Ingo

-------------->
commit d94b43b196712688c7b018edd69a4ca1650f61e0

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:55:10 2008 +0100

fix warning in drivers/net/s2io.c

this warning:

drivers/net/s2io.c: In function ‘rx_intr_handler’:
drivers/net/s2io.c:7369: warning: ‘lro’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between:

- s2io_club_tcp_session()and 'lro'

Annotate it.

index 6a1375f..a150575 100644
@@ -7366,7 +7366,7 @@ static int rx_osm_handler(struct ring_info *ring_data, struct RxD_t * rxdp)
 	int ring_no = ring_data->ring_no;
 	u16 l3_csum, l4_csum;
 	unsigned long long err = rxdp->Control_1 & RXD_T_CODE;
-	struct lro *lro;
+	struct lro *uninitialized_var(lro);
 	u8 err_mask;
 
 	skb->dev = dev;

commit 78f065dd3b7f8116dafc7eb6bf966e8276580c15

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:58:10 2008 +0100

fix warnings in drivers/net/niu.c

these warnings:

drivers/net/niu.c: In function ‘serdes_init_niu_1g_serdes’:
drivers/net/niu.c:451: warning: ‘sig’ may be used uninitialized in this function drivers/net/niu.c: In function ‘serdes_init_niu_10g_serdes’:
drivers/net/niu.c:550: warning: ‘sig’ may be used uninitialized in this function

triggers because GCC does not recognize that the max_retry loop always initializes 'sig', due to max_retry != 0.

Annotate them.

index afa399c..7e25142 100644
@@ -448,7 +448,7 @@ static int serdes_init_niu_1g_serdes(struct niu *np)
 	struct niu_link_config *lp = &np->link_config;
 	u16 pll_cfg, pll_sts;
 	int max_retry = 100;
-	u64 sig, mask, val;
+	u64 uninitialized_var(sig), mask, val;
 	u32 tx_cfg, rx_cfg;
 	unsigned long i;
 	int err;

@@ -547,7 +547,7 @@ static int serdes_init_niu_10g_serdes(struct niu *np)
 	struct niu_link_config *lp = &np->link_config;
 	u32 tx_cfg, rx_cfg, pll_cfg, pll_sts;
 	int max_retry = 100;
-	u64 sig, mask, val;
+	u64 uninitialized_var(sig), mask, val;
 	unsigned long i;
 	int err;
 

commit 9cc0c65af74e02da2c3d621cb061bd61b64cc24e

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:54:50 2008 +0100

fix warning in drivers/net/niu.c

this warning:

drivers/net/niu.c: In function ‘esr_reset’:
drivers/net/niu.c:741: warning: ‘reset’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between:

- esr_read_reset() and 'reset'

Annotate it.

index 1b6f548..afa399c 100644
@@ -738,7 +738,7 @@ static int esr_write_glue0(struct niu *np, unsigned long chan, u32 val)
 
 static int esr_reset(struct niu *np)
 {
-	u32 reset;
+	u32 uninitialized_var(reset);
 	int err;
 
 	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,

commit 32fd3b535aa43e290c33873124f91fec6a4329f7

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:37:40 2008 +0100

fix warning in drivers/net/qla3xxx.c

this warning:

drivers/net/qla3xxx.c: In function ‘ql3xxx_probe’:
drivers/net/qla3xxx.c:3912: warning: ‘pci_using_dac’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between 'pci_using_dac' and 'err'.

Annotate it.

index 508452c..3ae623e 100644
@@ -3909,7 +3909,7 @@ static int __devinit ql3xxx_probe(struct pci_dev *pdev,
 	struct net_device *ndev = NULL;
 	struct ql3_adapter *qdev = NULL;
 	static int cards_found = 0;
-	int pci_using_dac, err;
+	int uninitialized_var(pci_using_dac), err;
 
 	err = pci_enable_device(pdev);
 	if (err) {

commit ba58621694ef08b59b753bc5042b3db9472c5aff

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:03:21 2008 +0100

fix warning in net/sunrpc/xprtrdma/svc_rdma_transport.c

this warning:

net/sunrpc/xprtrdma/svc_rdma_transport.c: In function ‘svc_rdma_accept’:
net/sunrpc/xprtrdma/svc_rdma_transport.c:830: warning: ‘dma_mr_acc’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) flow connection between need_dma_mr and dma_mr_acc.

Annotate it.

index 6fb493c..636095c 100644
@@ -827,7 +827,7 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	struct rdma_conn_param conn_param;
 	struct ib_qp_init_attr qp_attr;
 	struct ib_device_attr devattr;
-	int dma_mr_acc;
+	int uninitialized_var(dma_mr_acc);
 	int need_dma_mr;
 	int ret;
 	int i;

commit b91eb990a0d839e810bb15048af08a69ab1b6ffe

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 12:03:12 2008 +0100

fix warning in net/sched/sch_hfsc.c

this warning:

net/sched/sch_hfsc.c: In function ‘hfsc_enqueue’:
net/sched/sch_hfsc.c:1577: warning: ‘err’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between hfsc_classify(), 'cl' and 'err'.

Annotate it.

index c1e77da..b39e3b5 100644
@@ -1574,7 +1574,7 @@ static int
 hfsc_enqueue(struct sk_buff *skb, struct Qdisc *sch)
 {
 	struct hfsc_class *cl;
-	int err;
+	int uninitialized_var(err);
 
 	cl = hfsc_classify(skb, sch, &err);
 	if (cl == NULL) {

commit c390a3784564c1fb2183e4a31ea39a9beb47d034

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:31:28 2008 +0100

fix warning in net/netfilter/nf_conntrack_ftp.c

this warning:

net/netfilter/nf_conntrack_ftp.c: In function ‘help’:
net/netfilter/nf_conntrack_ftp.c:360: warning: ‘matchoff’ may be used uninitialized in this function net/netfilter/nf_conntrack_ftp.c:360: warning: ‘matchlen’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between find_pattern(), 'found', 'matchoff' and 'matchlen'.

Annotate it.

index 4f71071..af7ad91 100644
@@ -357,7 +357,7 @@ static int help(struct sk_buff *skb,
 	int ret;
 	u32 seq;
 	int dir = CTINFO2DIR(ctinfo);
-	unsigned int matchlen, matchoff;
+	unsigned int uninitialized_var(matchlen), uninitialized_var(matchoff);
 	struct nf_ct_ftp_master *ct_ftp_info = &nfct_help(ct)->help.ct_ftp_info;
 	struct nf_conntrack_expect *exp;
 	union nf_inet_addr *daddr;

@@ -592,3 +592,4 @@ static int __init nf_conntrack_ftp_init(void)
 
 module_init(nf_conntrack_ftp_init);
 module_exit(nf_conntrack_ftp_fini);
+

commit 8fecea67aa434dd1653c4a275532e29903caeb88

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:31:20 2008 +0100

fix warning in net/ipv6/ip6_flowlabel.c

this warning:

net/ipv6/ip6_flowlabel.c: In function ‘ipv6_flowlabel_opt’:
net/ipv6/ip6_flowlabel.c:467: warning: ‘err’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between fl_create() and 'err'.

Annotate it.

index 37a4e77..62809ec 100644
@@ -464,7 +464,7 @@ static inline void fl_link(struct ipv6_pinfo *np, struct ipv6_fl_socklist *sfl,
 
 int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
 {
-	int err;
+	int uninitialized_var(err);
 	struct net *net = sock_net(sk);
 	struct ipv6_pinfo *np = inet6_sk(sk);
 	struct in6_flowlabel_req freq;

commit 67230f214ec140d07fcd32535841ee6eb4437364

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:31:12 2008 +0100

fix warning in net/dsa/mv88e6xxx.c

this warning:

net/dsa/mv88e6xxx.c: In function ‘mv88e6xxx_poll_link’:
net/dsa/mv88e6xxx.c:361: warning: ‘port_status’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between 'link' and 'port_status'.

Annotate it.

index aa6c609..4e4d8b5 100644
@@ -358,7 +358,7 @@ void mv88e6xxx_poll_link(struct dsa_switch *ds)
 
 	for (i = 0; i < DSA_MAX_PORTS; i++) {
 		struct net_device *dev;
-		int port_status;
+		int uninitialized_var(port_status);
 		int link;
 		int speed;
 		int duplex;

commit 948158e35984f86e565db303e0f026e536edca85

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:31:04 2008 +0100

fix warning in net/dsa/mv88e6060.c

this warning:

net/dsa/mv88e6060.c: In function ‘mv88e6060_poll_link’:
net/dsa/mv88e6060.c:225: warning: ‘port_status’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between 'link' and 'port_status'.

Annotate it.

index 54068ef..bfb4a64 100644
@@ -222,7 +222,7 @@ static void mv88e6060_poll_link(struct dsa_switch *ds)
 
 	for (i = 0; i < DSA_MAX_PORTS; i++) {
 		struct net_device *dev;
-		int port_status;
+		int uninitialized_var(port_status);
 		int link;
 		int speed;
 		int duplex;

commit 8574b3b4834c740d37d6eb2d6839037779b56da4

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:30:57 2008 +0100

fix warning in fs/dlm/netlink.c

this warning:

fs/dlm/netlink.c: In function ‘dlm_timeout_warn’:
fs/dlm/netlink.c:131: warning: ‘send_skb’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between prepare_data() and send_skb.

Annotate it.

index 18bda83..aa2a577 100644
@@ -127,8 +127,8 @@ static void fill_data(struct dlm_lock_data *data, struct dlm_lkb *lkb)
 
 void dlm_timeout_warn(struct dlm_lkb *lkb)
 {
+	struct sk_buff *uninitialized_var(send_skb);
 	struct dlm_lock_data *data;
-	struct sk_buff *send_skb;
 	size_t size;
 	int rv;
 

commit 9b9c8fde6c9d175bb7a108e3d231626984364451

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 11:30:15 2008 +0100

fix warning in drivers/net/sis900.c

this warning:

drivers/net/sis900.c: In function ‘sis900_timer’:
drivers/net/sis900.c:1280: warning: ‘speed’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between sis900_read_mode(), 'speed' and 'duplex'.

Annotate it.

index fa3a460..8be433e 100644
@@ -1277,7 +1277,7 @@ static void sis900_timer(unsigned long data)
 	u16 status;
 
 	if (!sis_priv->autong_complete){
-		int speed, duplex = 0;
+		int uninitialized_var(speed), duplex = 0;
 
 		sis900_read_mode(net_dev, &speed, &duplex);
 		if (duplex){

commit 924a289446dadc9c3c9896c87446e0403c746537

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:38:24 2008 +0100

fix warning in drivers/net/wireless/iwlwifi/iwl-3945.c

this warning:

drivers/net/wireless/iwlwifi/iwl-3945.c: In function ‘iwl3945_txpower_set_from_eeprom’:
drivers/net/wireless/iwlwifi/iwl-3945.c:2182: warning: ‘power_idx’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between iwl3945_hw_reg_get_matched_power_index() and power_idx.

Annotate it.

index 7ca5627..e9e0b1a 100644
@@ -2179,7 +2179,7 @@ int iwl3945_txpower_set_from_eeprom(struct iwl3945_priv *priv)
 		/* set tx power value for all OFDM rates */
 		for (rate_index = 0; rate_index < IWL_OFDM_RATES;
 		     rate_index++) {
-			s32 power_idx;
+			s32 uninitialized_var(power_idx);
 			int rc;
 
 			/* use channel group's clip-power table,

commit 22672393d84dc3c74b7cd46eebdb44d43747318b

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:38:08 2008 +0100

fix warning in drivers/net/wireless/b43/main.c

this warning:

drivers/net/wireless/b43/main.c: In function ‘b43_op_config’:
drivers/net/wireless/b43/main.c:3258: warning: ‘gmode’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between the list_for_each_entry() loop, up_dev and gmode.

Annotate it.

index 14c44df..bd70b27 100644
@@ -3255,7 +3255,7 @@ static int b43_switch_band(struct b43_wl *wl, struct ieee80211_channel *chan)
 	struct b43_wldev *down_dev;
 	struct b43_wldev *d;
 	int err;
-	bool gmode;
+	bool uninitialized_var(gmode);
 	int prev_status;
 
 	/* Find a device and PHY which supports the band. */

commit de965062d615afc885bd800976e488bd3c2d1d6f

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:37:58 2008 +0100

fix warning in drivers/net/wireless/ath9k/rc.c

this warning:

drivers/net/wireless/ath9k/rc.c: In function ‘ath_rc_rate_getidx’:
drivers/net/wireless/ath9k/rc.c:1033: warning: ‘nextindex’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between ath_rc_get_nextlowervalid_txrate() and &nextindex.

Annotate it.

index cca2fc5..27b35d3 100644
@@ -1030,7 +1030,7 @@ static u8 ath_rc_rate_getidx(struct ath_softc *sc,
 			     u16 min_rate)
 {
 	u32 j;
-	u8 nextindex;
+	u8 uninitialized_var(nextindex);
 	struct ath_tx_ratectrl *rate_ctrl =
 		(struct ath_tx_ratectrl *)(ath_rc_priv);
 

commit e5e5492e7094ed4bea8daf0302f60fbb21daa4c7

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 14:52:20 2008 +0100

fix warning in drivers/net/wan/z85230.c

this warning:

drivers/net/wan/z85230.c: In function ‘z8530_interrupt’:
drivers/net/wan/z85230.c:713: warning: ‘intr’ may be used uninitialized in this function

is clearly bogus - annotate it.

index 5bf7e01..d7bf53a 100644
@@ -710,7 +710,7 @@ EXPORT_SYMBOL(z8530_nop);
 irqreturn_t z8530_interrupt(int irq, void *dev_id)
 {
 	struct z8530_dev *dev=dev_id;
-	u8 intr;
+	u8 uninitialized_var(intr);
 	static volatile int locker=0;
 	int work=0;
 	struct z8530_irqhandler *irqs;

commit 742cddc749294be3c0fa909e6255c5a338a1ac40

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:37:47 2008 +0100

fix error path in drivers/net/mlx4/en_rx.c

this warning:

drivers/net/mlx4/en_rx.c: In function ‘mlx4_en_activate_rx_rings’:
drivers/net/mlx4/en_rx.c:412: warning: ‘err’ may be used uninitialized in this function

Triggers because 'err' is uninitialized in the following input
conditions: priv->rx_ring_num is zero and mlx4_en_fill_rx_buffers() fails.

But even if ->rx_ring_num is nonzero, 'err' will be zero if mlx4_en_fill_rx_buffers() fails and mlx4_en_activate_rx_rings() returns success - incorrectly.

So it's best to keep the error code uptodate on mlx4_en_fill_rx_buffers() calls as well.

index 6232227..aed7739 100644
@@ -443,7 +443,8 @@ int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv)
 		/* Fill Rx buffers */
 		ring->full = 0;
 	}
-	if (mlx4_en_fill_rx_buffers(priv))
+	err = mlx4_en_fill_rx_buffers(priv);
+	if (err)
 		goto err_buffers;
 
 	for (ring_ind = 0; ring_ind < priv->rx_ring_num; ring_ind++) {

commit 181787d1c30af364ffd15428fe6fab7798c3f3c1

Author: Ingo Molnar <[email protected]>
Date:   Tue Nov 25 10:37:35 2008 +0100

fix warning in drivers/net/cassini.c

this warning:

drivers/net/cassini.c: In function ‘cas_rx_ringN’:
drivers/net/cassini.c:2350: warning: ‘skb’ may be used uninitialized in this function

triggers because GCC does not recognize the (correct) error flow between cas_rx_process_pkt() and 'skb'.

Annotate it.

index 86909cf..e083e1c 100644
@@ -2347,7 +2347,7 @@ static int cas_rx_ringN(struct cas *cp, int ring, int budget)
 	drops = 0;
 	while (1) {
 		struct cas_rx_comp *rxc = rxcs + entry;
-		struct sk_buff *skb;
+		struct sk_buff *uninitialized_var(skb);
 		int type, len;
 		u64 words[4];
 		int i, dring;

commit 96c79254264033f1505198d5d362b68d88d1bbaf

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 12:41:30 2008 +0200

fix warning in net/dccp/options.c

this warning:

net/dccp/options.c: In function ‘dccp_parse_options’:
net/dccp/options.c:67: warning: ‘value’ may be used uninitialized in this function

is a bogus GCC warning. The compiler does not recognize the relation between "value" and "mandatory" variables: the code flow can ever reach the "out_invalid_option:" label if 'mandatory' is set to 1, and when 'mandatory' is non-zero, we'll always have 'value' initialized.

Help out the compiler by annotating the variable.

index 0809b63..18dcfb9 100644
@@ -64,7 +64,7 @@ int dccp_parse_options(struct sock *sk, struct dccp_request_sock *dreq,
 					(dh->dccph_doff * 4);
 	struct dccp_options_received *opt_recv = &dp->dccps_options_received;
 	unsigned char opt, len;
-	unsigned char *value;
+	unsigned char *uninitialized_var(value);
 	u32 elapsed_time;
 	__be32 opt_val;
 	int rc;

commit 193e56a10e56505dc7eac9e5632834c72b55613d

Author: Ingo Molnar <[email protected]>
Date:   Mon Aug 18 16:18:34 2008 +0200

fix warning in drivers/net/mlx4/mcg.c

fix warning:

drivers/net/mlx4/mcg.c: In function ‘mlx4_multicast_attach’:
drivers/net/mlx4/mcg.c:217: warning: integer overflow in expression

there was no real danger of overflow here though.

md5:
db8eb55620f886c03854a2abb2ce6c3f mcg.o.before.asm db8eb55620f886c03854a2abb2ce6c3f mcg.o.after.asm

index 592c01a..224ef0f 100644
@@ -215,7 +215,7 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
 
 	if (block_mcast_loopback)
 		mgm->qp[members_count++] = cpu_to_be32((qp->qpn & MGM_QPN_MASK) |
-						       (1 << MGM_BLCK_LB_BIT));
+						       (1U << MGM_BLCK_LB_BIT));
 	else
 		mgm->qp[members_count++] = cpu_to_be32(qp->qpn & MGM_QPN_MASK);
 

commit bcc35f75a3b9b96c53a74176b135304fc7ffe052

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 15:49:15 2008 +0200

fix warning in net/ax25/sysctl_net_ax25.c

fix this warning:

net/ax25/sysctl_net_ax25.c:27: warning: ‘min_ds_timeout’ defined but not used net/ax25/sysctl_net_ax25.c:27: warning: ‘max_ds_timeout’ defined but not used

These are only used in the CONFIG_AX25_DAMA_SLAVE case.

index f288fc4..735ceef 100644
@@ -24,7 +24,9 @@ static int min_idle[1],			max_idle[] = {65535000};
 static int min_n2[] = {1},		max_n2[] = {31};
 static int min_paclen[] = {1},		max_paclen[] = {512};
 static int min_proto[1],		max_proto[] = { AX25_PROTO_MAX };
+#ifdef CONFIG_AX25_DAMA_SLAVE
 static int min_ds_timeout[1],		max_ds_timeout[] = {65535000};
+#endif
 
 static struct ctl_table_header *ax25_table_header;
 

commit 59681c6c30e3f700af8d716f66e41ad18f726b80

Author: Ingo Molnar <[email protected]>
Date:   Mon Oct 20 18:50:11 2008 +0200

fix warning in net/sunrpc/xprtrdma/verbs.c

fix this warning:

net/sunrpc/xprtrdma/verbs.c: In function ‘rpcrdma_conn_upcall’:
net/sunrpc/xprtrdma/verbs.c:279: warning: unused variable ‘addr’

index a5fef5e..900e595 100644
@@ -276,7 +276,9 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 	struct rpcrdma_xprt *xprt = id->context;
 	struct rpcrdma_ia *ia = &xprt->rx_ia;
 	struct rpcrdma_ep *ep = &xprt->rx_ep;
+#ifdef RPC_DEBUG
 	struct sockaddr_in *addr = (struct sockaddr_in *) &ep->rep_remote_addr;
+#endif
 	struct ib_qp_attr attr;
 	struct ib_qp_init_attr iattr;
 	int connstate = 0;

commit ad8d03b315507a954160fcf04dc94cef9f118e77

Author: Ingo Molnar <[email protected]>
Date:   Mon Oct 20 17:34:02 2008 +0200

fix warning in net/bluetooth/rfcomm/sock.c

fix this warning:

net/bluetooth/rfcomm/sock.c: In function ‘rfcomm_sock_ioctl’:
net/bluetooth/rfcomm/sock.c:795: warning: unused variable ‘sk’

perhaps BT_DEBUG() should be improved to do printf format checking instead of the #ifdef, but that looks quite intrusive: each bluetooth .c file undefines the macro.

index 8a972b6..bc0d4a7 100644
@@ -792,7 +792,9 @@ static int rfcomm_sock_getsockopt(struct socket *sock, int level, int optname, c
 
 static int rfcomm_sock_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 {
+#if defined(CONFIG_BT_RFCOMM_TTY) || defined(CONFIG_BT_RFCOMM_DEBUG)
 	struct sock *sk = sock->sk;
+#endif
 	int err;
 
 	BT_DBG("sk %p cmd %x arg %lx", sk, cmd, arg);

commit 320451a3f79098ca800f2e58674074c26efee020

Author: Ingo Molnar <[email protected]>
Date:   Sat Oct 18 21:33:06 2008 +0200

fix warning in drivers/net/wireless/ray_cs.c

fix this warning:

drivers/net/wireless/ray_cs.c:297: warning: ‘rcsid’ defined but not used

'rcsid' is only used in the CONFIG_PROC_FS=y case.

index 1404a57..c23a811 100644
@@ -294,7 +294,9 @@ static char hop_pattern_length[] = { 1,
 	     JAPAN_TEST_HOP_MOD
 };
 
+#ifdef CONFIG_PROC_FS
 static char rcsid[] = "Raylink/WebGear wireless LAN - Corey <Thomas [email protected]>";
+#endif
 
 /*=============================================================================
     ray_attach() creates an "instance" of the driver, allocating

commit decc1656e870c9e5bd94f2357b8496b8da6c2e73

Author: Ingo Molnar <[email protected]>
Date:   Sat Oct 18 19:14:40 2008 +0200

fix warning in net/netfilter/nf_conntrack_proto_sctp.c

fix this warning:

net/netfilter/nf_conntrack_proto_sctp.c: In function ‘sctp_packet’:
net/netfilter/nf_conntrack_proto_sctp.c:376: warning: array subscript is above array bounds

GCC noticed that new_state could in theory be SCTP_CONNTRACK_MAX, which is beyond the limit of the sctp_timeouts[] array.

This situation could happen if for_each_sctp_chunk() exits immediately. That should not normally happen - but it's not obvious to me and it's better to insert an extra check to the end of the loop for this case.

This gets rid of warning as well.

index ae8c260..d31ced4 100644
@@ -373,6 +373,9 @@ static int sctp_packet(struct nf_conn *ct,
 	}
 	write_unlock_bh(&sctp_lock);
 
+	if (new_state == SCTP_CONNTRACK_MAX)
+		goto out;
+
 	nf_ct_refresh_acct(ct, ctinfo, skb, sctp_timeouts[new_state]);
 
 	if (old_state == SCTP_CONNTRACK_COOKIE_ECHOED &&

commit b547d51a26c5ed23571df50c7254439e2f5dbebe

Author: Ingo Molnar <[email protected]>
Date:   Sat Oct 18 19:02:37 2008 +0200

fix warning in net/netfilter/ipvs/ip_vs_ctl.c

Impact: fix potential kernel stack information exposure to user-space.

fix this warning:

net/netfilter/ipvs/ip_vs_ctl.c: In function ‘ip_vs_genl_set_cmd’:
net/netfilter/ipvs/ip_vs_ctl.c:3034: warning: ‘t.udp_timeout’ may be used uninitialized in this function net/netfilter/ipvs/ip_vs_ctl.c:3034: note: ‘t.udp_timeout’ was declared here

GCC seems to be correct here: in the !CONFIG_IP_VS_PROTO_UDP case __ip_vs_get_timeouts keeps the timeout structure uninitialized. But in a couple of places we use and copy this structure to user-space, potentially exposing kernel-stack data to user-space.

While it's a minor issue that only triggers with certain config variations, it's best to always initialize this structure in __ip_vs_get_timeouts(). (GCC will eliminate this memset)

index 0302cf3..7e3a0c7 100644
@@ -2327,6 +2327,7 @@ __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
 static inline void
 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
 {
+	memset(u, 0, sizeof(*u));
 #ifdef CONFIG_IP_VS_PROTO_TCP
 	u->tcp_timeout =
 		ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;

commit 2281d27663a5f17689ec49e029b4962cba19d614

Author: Ingo Molnar <[email protected]>
Date:   Sat Oct 18 16:27:37 2008 +0200

fix warning in drivers/net/atlx/atl2.c

fix this warning:

drivers/net/atlx/atl2.c: In function ‘atl2_request_irq’:
drivers/net/atlx/atl2.c:644: warning: unused variable ‘err’

'err' is unused in the !CONFIG_PCI_MSI case.

Instead of further increasing the #ifdeffery in this function, restructure the code a bit and get rid of the #ifdef. This relies on the fact that pci_enable_msi() will always fail in the !CONFIG_PCI_MSI case.

There should be no change in driver behavior.

index 8571e8c..e02bd7e 100644
@@ -644,7 +644,6 @@ static int atl2_request_irq(struct atl2_adapter *adapter)
 	int flags, err = 0;
 
 	flags = IRQF_SHARED;
-#ifdef CONFIG_PCI_MSI
 	adapter->have_msi = true;
 	err = pci_enable_msi(adapter->pdev);
 	if (err)

@@ -652,7 +651,6 @@ static int atl2_request_irq(struct atl2_adapter *adapter)
 
 	if (adapter->have_msi)
 		flags &= ~IRQF_SHARED;
-#endif
 
 	return request_irq(adapter->pdev->irq, &atl2_intr, flags, netdev->name,
 		netdev);

commit b5a7c2b9e1a1b4d80da5d66b330545fdeafe1dad

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 19:48:51 2008 +0200

fix warning in net/netfilter/nf_conntrack_proto_tcp.c

fix this warning:

net/netfilter/nf_conntrack_proto_tcp.c: In function ‘tcp_in_window’:
net/netfilter/nf_conntrack_proto_tcp.c:491: warning: unused variable ‘net’ net/netfilter/nf_conntrack_proto_tcp.c: In function ‘tcp_packet’:
net/netfilter/nf_conntrack_proto_tcp.c:812: warning: unused variable ‘net’

index 7f2f43c..debdaf7 100644
@@ -129,7 +129,7 @@ extern const struct nla_policy nf_ct_port_nla_policy[];
 	 && net_ratelimit())
 #endif
 #else
-#define LOG_INVALID(net, proto) 0
+static inline int LOG_INVALID(struct net *net, int proto) { return 0; }
 #endif /* CONFIG_SYSCTL */
 
 #endif /*_NF_CONNTRACK_PROTOCOL_H*/

commit 822c18db871bc88b2a60d7a9618ec79a8b5e1ef5

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 18:31:33 2008 +0200

fix warning in drivers/net/wireless/iwlwifi/iwl-scan.c

fix this warning:

drivers/net/wireless/iwlwifi/iwl-scan.c:92: warning: ‘iwl_escape_essid’ defined but not used

Convert the IWL debug macros to proper printf format checking. Besides better type checking, this gets rid of the warning as well.

index e548d67..ddc1fbf 100644
@@ -68,8 +68,10 @@ void iwl_dbgfs_unregister(struct iwl_priv *priv);
 #endif
 
 #else
-#define IWL_DEBUG(level, fmt, args...)
-#define IWL_DEBUG_LIMIT(level, fmt, args...)
+static inline void __attribute__ ((format(printf, 2, 3)))
+IWL_DEBUG(int level, const char *fmt, ...) { }
+static inline void __attribute__ ((format(printf, 2, 3)))
+IWL_DEBUG_LIMIT(int level, const char *fmt, ...) { }
 #endif				/* CONFIG_IWLWIFI_DEBUG */
 
 

index 4eee1b1..f4f0664 100644
@@ -43,7 +43,6 @@
 #include "iwl-io.h"
 #include "iwl-helpers.h"
 
-#ifdef CONFIG_IWLWIFI_DEBUG
 static const char *led_type_str[] = {
 	__stringify(IWL_LED_TRG_TX),
 	__stringify(IWL_LED_TRG_RX),

@@ -51,7 +50,6 @@ static const char *led_type_str[] = {
 	__stringify(IWL_LED_TRG_RADIO),
 	NULL
 };
-#endif /* CONFIG_IWLWIFI_DEBUG */
 
 
 static const struct {

commit 69f38f617c3f832e088be9d35242d929aa623fd4

Author: Ingo Molnar <[email protected]>
Date:   Wed Aug 27 19:06:48 2008 +0200

fix warning in drivers/net/depca.c

fix warning:

drivers/net/depca.c: In function ‘depca_eisa_probe’:
drivers/net/depca.c:1564: warning: ‘mem_start’ may be used uninitialized in this function

this seems to be a real bug - depca_eisa_probe() does not check for failure. Add it, symmetric to depca_isa_probe().

index ace39ec..f00bd64 100644
@@ -1558,6 +1558,7 @@ static int __init depca_isa_probe (struct platform_device *device)
 #ifdef CONFIG_EISA
 static int __init depca_eisa_probe (struct device *device)
 {
+	enum depca_type adapter = unknown;
 	struct eisa_device *edev;
 	struct net_device *dev;
 	struct depca_private *lp;

@@ -1576,7 +1577,11 @@ static int __init depca_eisa_probe (struct device *device)
 	 * the EISA configuration structures (yet... :-), just rely on
 	 * the ISA probing to sort it out... */
 
-	depca_shmem_probe (&mem_start);
+	adapter = depca_shmem_probe (&mem_start);
+	if (adapter == unknown) {
+		status = -ENODEV;
+		goto out_free;
+	}
 
 	dev->base_addr = ioaddr;
 	dev->irq = irq;

commit 557073e3dac7b8479df69c140a5fd2b8f4a9cfad

Author: Ingo Molnar <[email protected]>
Date:   Wed Aug 27 19:05:19 2008 +0200

fix warning in drivers/net/3c523.c

fix warning:

drivers/net/3c523.c:582: warning: ‘cleanup_card’ defined but not used

No code changed:

md5:
ebe4a1b27d3f21b0b12a78c58463b0d7 3c523.o.before.asm ebe4a1b27d3f21b0b12a78c58463b0d7 3c523.o.after.asm

index e2ce41d..7b9d427 100644
@@ -578,13 +578,13 @@ err_out:
 	return retval;
 }
 
+#ifdef MODULE
 static void cleanup_card(struct net_device *dev)
 {
 	mca_set_adapter_procfn(((struct priv *) (dev->priv))->slot, NULL, NULL);
 	release_region(dev->base_addr, ELMC_IO_EXTENT);
 }
-
-#ifndef MODULE
+#else
 struct net_device * __init elmc_probe(int unit)
 {
 	struct net_device *dev = alloc_etherdev(sizeof(struct priv));

commit d138e44e2f8d0f26e04b0386d328d9cc7e33d82e

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 14:30:37 2008 +0200

drivers/net/wireless/b43/phy_g.c: type check debug printouts

this warning:

drivers/net/wireless/b43/phy_g.c: In function ‘b43_gphy_op_recalc_txpower’:
drivers/net/wireless/b43/phy_g.c:3191: warning: unused variable ‘dbm’

is caused because b43dbg() is a macro, hence in the !B43_DEBUG case the compiler does not know that the variables are used.

Convert it to a proper inline instead. This also improves type checking in the !B43_DEBUG case.

index 427b820..ac55b62 100644
@@ -853,7 +853,8 @@ void b43warn(struct b43_wl *wl, const char *fmt, ...)
 void b43dbg(struct b43_wl *wl, const char *fmt, ...)
     __attribute__ ((format(printf, 2, 3)));
 #else /* DEBUG */
-# define b43dbg(wl, fmt...) do { /* nothing */ } while (0)
+static inline void __attribute__ ((format(printf, 2, 3)))
+b43dbg(struct b43_wl *wl, const char *fmt, ...) { }
 #endif /* DEBUG */
 
 /* A WARN_ON variant that vanishes when b43 debugging is disabled.

commit 7e4f7889aca92f2cf95ba31e005fc1368c1cbc5a

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 17:47:04 2008 +0200

fix warning in drivers/net/sky2.c

fix this warning:

drivers/net/sky2.c: In function ‘sky2_mhz’:
drivers/net/sky2.c:2753: warning: control reaches end of non-void function

BUG() can be NOP on !CONFIG_BUG.

index 3813d15..85fdacc 100644
@@ -2748,7 +2748,7 @@ static u32 sky2_mhz(const struct sky2_hw *hw)
 		return 156;
 
 	default:
-		BUG();
+		panic("sky2_mhz: unknown chip id!");
 	}
 }
 

commit deb797f96d920083ba8a0bcfa6f57b19fc593571

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 17:34:27 2008 +0200

fix warning in net/sunrpc/svcauth_unix.c

fix this warning:

net/sunrpc/svcauth_unix.c: In function ‘svcauth_unix_set_client’:
net/sunrpc/svcauth_unix.c:681: warning: ‘sin6’ may be used uninitialized in this function

caused by BUG() being a NOP on !CONFIG_BUG.

index f24800f..200ed93 100644
@@ -692,7 +692,7 @@ svcauth_unix_set_client(struct svc_rqst *rqstp)
 		sin6 = svc_addr_in6(rqstp);
 		break;
 	default:
-		BUG();
+		panic("svcauth_unix_set_client: bad address family!");
 	}
 
 	rqstp->rq_client = NULL;

@@ -873,3 +873,4 @@ struct auth_ops svcauth_unix = {
 	.set_client	= svcauth_unix_set_client,
 };
 
+

commit 923a5a960577dcc5189b6113300a122c0212b292

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 16:55:34 2008 +0200

fix warning in security/selinux/netnode.c

fix this warning:

security/selinux/netnode.c: In function ‘sel_netnode_find’:
security/selinux/netnode.c:131: warning: ‘idx’ may be used uninitialized in this function

BUG() can be a NOP in the !CONFIG_BUG case. Return NULL in that case, instead of running in an undefined way.

index 7100072..a307122 100644
@@ -140,6 +140,7 @@ static struct sel_netnode *sel_netnode_find(const void *addr, u16 family)
 		break;
 	default:
 		BUG();
+		return NULL;
 	}
 
 	list_for_each_entry_rcu(node, &sel_netnode_hash[idx].list, list)

commit c62cda5dfaf7630edefe097a3e23af7732060b08

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 16:26:02 2008 +0200

fix warning in net/packet/af_packet.c

fix this warning:

net/packet/af_packet.c: In function ‘tpacket_rcv’:
net/packet/af_packet.c:604: warning: ‘hdrlen’ may be used uninitialized in this function

This is because in the !CONFIG_BUG case BUG() is a NOP.

There is no other sane way to continue but to panic the system unconditionally, regardless of the CONFIG_BUG setting.

index c718e7e..1478fb8 100644
@@ -709,7 +709,7 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, struct packe
 		hdrlen = sizeof(*h.h2);
 		break;
 	default:
-		BUG();
+		panic("AF_PACKET: bad tp->version");
 	}
 
 	sll = h.raw + TPACKET_ALIGN(hdrlen);

commit d581867031ebe3c6f4e018ea32ed8b9ee5141ead

Author: Ingo Molnar <[email protected]>
Date:   Fri Oct 17 16:15:59 2008 +0200

fix warning in net/rfkill/rfkill.c

fix this warning:

net/rfkill/rfkill.c: In function ‘rfkill_get_type_str’:
net/rfkill/rfkill.c:393: warning: control reaches end of non-void function

in the embedded !CONFIG_BUG case BUG() is a NOP.

Return NULL to solve this (unlikely) scenario. All callers feed this pointer into sprintf %s so NULL should be fine.

index f949a48..eecb9a8 100644
@@ -389,6 +389,7 @@ static const char *rfkill_get_type_str(enum rfkill_type type)
 		return "wwan";
 	default:
 		BUG();
+		return NULL;
 	}
 }
 

--
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

I already applied these patches, see my reply in that thread.

I was saying to submit these (individually, please!) next time with netdev CC:'d.

Posting a ton all at once is pointless, they all get logged to patchwork as one big blog and I cannot annotate them individually and give you proper feedback.

From: Ingo Molnar <[email protected]>
Date: Tue, 25 Nov 2008 18:14:38 +0100

Applied.

Patrick got this one.

Applied

Needs to be submitted via wireless folks at linux-wireless

Applied.

Submit to wireless maintainer please.

Patrick got these.

Applied.

Patrick got it.

Wireless submission please.

Applied.

Submit to wireless folks please.

Not applied. Changing this from a BUG() to a panic() is wrong. BUG() kills the current execution context, and allows the system to potentially continue. Panic() takes down the entire machine.

The !CONFIG_BUG BUG() definition should be marked in such a way that gcc thinks it is noreturn. Otherwise BUG() isn't really BUG() anymore. The other option is to never allow BUG to be disabled or have it always evaluate to a simple bug trap.

Yeah, although in practice a BUG() in some of those places will take down the whole machine:

That's packet receive path, holding spinlocks or in other atomic context (irq/softirq)? A BUG() will take down the machine as we will crash and try to schedule in atomic context.

Yeah, but note that that's exactly the current behavior (and intent) of !CONFIG_BUG: it is not doing anything (it's just an empty macro which returns!), and we get those build warnings because the control flow becomes undefined. We cannot mark it noreturn because it does return.

Ingo

Hi Linus:

At this late stage there is just one fix for a regression from 2.6.27. We used to allow algorithms to be built-in with the new testing framework built as a module. This meant that the built-in algorithms may not get tested properly if modules can't be loaded when the built-in algorithms are registered. It also seems to trigger a modprobe loop on one distro. This configuration is now forbidden.

Please pull from

git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

or

master.kernel.org:/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

Herbert Xu (1):
crypto: api - Disallow cryptomgr as a module if algorithms are built-in

 crypto/Kconfig  |   36 ++++++++++++++++++++++++++++++++----
 crypto/Makefile |   18 +++++++++---------
 2 files changed, 41 insertions(+), 13 deletions(-)

Thanks,

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Hi Linus:

A regression has been reported where the new algorithm testing infrastructure may cause the optimised versions of AES to fail when it's built into the kernel (as opposed to as a module). This is due to its use of shared AES tables that may not have been initialised yet. This has been fixed by pregenerating the tables instead of having them computed at boot/run-time.

Please pull from

git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

or

master.kernel.org:/pub/scm/linux/kernel/git/herbert/crypto-2.6.git

Herbert Xu (1):
crypto: aes - Precompute tables

 crypto/aes_generic.c | 1145 ++++++++++++++++++++++++++++++++++++++++++++++----
 include/crypto/aes.h |    8 +-
 2 files changed, 1059 insertions(+), 94 deletions(-)

Thanks,

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Can't we just get rid of that stupid testing infrastructure, or make it a module-only thing? I refuse to pull a 1000+ line patch for something like this, especially this late. It's insane.

Linus

Well the testing infrastructure simply flagged a problem that was always there. Previously if something used AES early enough it would silently give the wrong result instead. Of course this is fairly unlikely since the number of AES users is limited.

Now that the testing infrastructure is available, we could just do nothing for 2.6.28 since the C AES version will always be there and work correctly.

Thanks,

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt