Prior to 2.6.19, dm-crypt passes write barrier requests through to the underlying device in the order received unmodified (the ordering).
In 2.6.19 and newer, these writes are now passed through a workqueue. On an SMP box running encrpytion for write requests in parallel, wouldn't this cause the requests to arrive to the device in a different order? A 2.6.22 patch from Milan Broz sets dm-crypt to disable barriers, but if this disabling is being handled at the filesystem level, is there still a chance of data becoming corrupt? I'm currently running an unpatched 2.6.20 kernel on an SMP machine testing the validity of some static files going back and forth through dm-crypt on an XFS filesystem mounted with nobarrier to be sure, and so far I'm not noticing anything, but this is hardly a good test. Suggestions or any other way to help out appreciated. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
