Review of an OpenCrypto patch

[Benno Rice]

Hi -security,

I work at EMC Isilon and one of our developers has found a race in opencyrpto and provided the attached patch to address it.

The situation as explained to me is that the crypto request queue and dequeue operate under CRYPTO_Q_LOCK, along with crypto_invoke and thus crypto processing. Meanwhile crypto_newsession (and thus all driver new session calls) operate under CRYPTO_DRIVER_LOCK.

This leads to a situation where resizing of the swcr_sessions array in swcr_newsession can interfere with the use of that array in swcr_process.

The attached patch protects the swcr_sessions array with a new rwlock.

Could somebody give this a look over and let me know if it’s commitable roughly as is or needs some work?

Cheers,
	Benno.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: patch-117508.3
Type: application/octet-stream
Size: 3808 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-security/attachments/20140115/64997980/attachment.obj>
-------------- next part --------------

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 841 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.freebsd.org/pipermail/freebsd-security/attachments/20140115/64997980/attachment.sig>