fix nvme-tcp and nvme-rdma controller reset hangs when using multipath

Christoph Hellwig hch at lst.de
Mon Mar 22 07:37:24 GMT 2021


Hi all,

this series attempts to fix the nvme queue freeze deadlock Sagi reported.
It needs a little help from the block layer so that we can avoid blocking
on the queue free percpu_ref without forcing the entire NOWAIT flag and
thus also not waiting for tag allocations.



More information about the Linux-nvme mailing list