Re: [Qemu-devel] [PATCH] block: init bs->io_base correctly to avoid locking

[Paolo Bonzini]

Il 10/01/2013 13:12, Peter Lieven ha scritto:
>> > 
>> > But perhaps we do not need to start a slice at all when iolimits are
>> > set.  That is, do
>> > 
>> >  bs->slice_start = bs->slice_end = bs->slice_time = 0;
>> > 
>> > or perhaps even nothing at all since bdrv_io_limits_disable should have
>> > written those exact values.
> Or it was set when the BlockDriverState was initialized.
> 
> I am not familiar enough with the io limits code to decide if not starting a 
> slice
> is also correct.

I haven't tested it, but if it works, I think it is better.

Think of it this way: it doesn't matter whether the first I/O operation
comes immediately after limits are set, or 10 seconds later.  In the
latter case, bdrv_exceed_io_limits will _already_ start a new slice.  It
is better to be consistent and always delay the start of the slice.

Paolo