On Thu, Sep 03, 2020 at 06:42:12PM +0300, Vladimir Sementsov-Ogievskiy wrote:
> But still, is it OK to do blk_drain holding the global mutex? Drain may take 
> a relatively long time, and vm is not responding due to global mutex locked 
> in cpu thread..

This is a fundamental problem in QEMU. Unfortunately I/O requests cannot
be canceled quickly (some protocols do not support cancelation at all).

There are code paths like device reset or monitor commands where drain
can hang for an unbounded amount of time :(.

Stefan

Attachment: signature.asc
Description: PGP signature

Reply via email to