In the Linux kernel, the following vulnerability has been resolved:
md/raid5: fix deadlock that raid5d() wait for itself to clear MDSBCHANGE_PENDING
Xiao reported that lvm2 test lvconvert-raid-takeover.sh can hang with small possibility, the root cause is exactly the same as commit bed9e27baf52 ("Revert "md/raid5: Wait for MDSBCHANGE_PENDING in raid5d"")
However, Dan reported another hang after that, and junxiao investigated the problem and found out that this is caused by plugged bio can't issue from raid5d().
Current implementation in raid5d() has a weird dependence:
1) mdcheckrecovery() from raid5d() must hold 'reconfigmutex' to clear MDSBCHANGEPENDING; 2) raid5d() handles IO in a deadloop, until all IO are issued; 3) IO from raid5d() must wait for MDSBCHANGE_PENDING to be cleared;
This behaviour is introduce before v2.6, and for consequence, if other context hold 'reconfigmutex', and mdcheckrecovery() can't update superblock, then raid5d() will waste one cpu 100% by the deadloop, until 'reconfig_mutex' is released.
Refer to the implementation from raid1 and raid10, fix this problem by skipping issue IO if MDSBCHANGEPENDING is still set after mdcheckrecovery(), daemon thread will be woken up when 'reconfigmutex' is released. Meanwhile, the hang problem will be fixed as well.