On Fri, 2019-04-05 at 05:43 +-0800, Ming Lei wrote: +AD4 On Thu, Apr 04, 2019 at 10:08:43AM -0700, Bart Van Assche wrote: +AD4 +AD4 blk+AF8-mq+AF8-try+AF8-issue+AF8-directly() can return BLK+AF8-STS+ACoAXw-RESOURCE for requests that +AD4 +AD4 have been queued. If that happens when blk+AF8-mq+AF8-try+AF8-issue+AF8-directly() is called +AD4 +AD4 by the dm-mpath driver then dm-mpath will try to resubmit a request that is +AD4 +AD4 already queued and a kernel crash follows. Since it is nontrivial to fix +AD4 +AD4 blk+AF8-mq+AF8-request+AF8-issue+AF8-directly(), revert the blk+AF8-mq+AF8-request+AF8-issue+AF8-directly() +AD4 +AD4 changes that went into kernel v5.0. +AD4 +AD4 +AD4 +AD4 This patch reverts the following commits: +AD4 +AD4 +ACo d6a51a97c0b2 (+ACI-blk-mq: replace and kill blk+AF8-mq+AF8-request+AF8-issue+AF8-directly+ACI) +ACM v5.0. +AD4 +AD4 +ACo 5b7a6f128aad (+ACI-blk-mq: issue directly with bypass 'false' in blk+AF8-mq+AF8-sched+AF8-insert+AF8-requests+ACI) +ACM v5.0. +AD4 +AD4 +ACo 7f556a44e61d (+ACI-blk-mq: refactor the code of issue request directly+ACI) +ACM v5.0. +AD4 +AD4 I am fine to revert them now. +AD4 +AD4 However, could you revert them one by one so that they can be reviewed +AD4 easily? Anyone who wants to verify this patch can do that by reverting the three commits mentioned above and by using diff to compare the revert result with this patch. Thanks, Bart.
