[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v2 qemu-trad] HVM: atomically access pointers in bufioreq handling

Jan Beulich writes ("[PATCH v2 qemu-trad] HVM: atomically access pointers in 
bufioreq handling"):
> The number of slots per page being 511 (i.e. not a power of two) means
> that the (32-bit) read and write indexes going beyond 2^32 will likely
> disturb operation. The hypervisor side gets I/O req server creation
> extended so we can indicate that we're using suitable atomic accesses
> where needed, allowing it to atomically canonicalize both pointers when
> both have gone through at least one cycle.
> The Xen side counterpart (which is not a functional prereq to this
> change, albeit the intention is for Xen to assume default servers
> always use suitable atomic accesses) went in already (commit
> b7007bc6f9).

Acked-by: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>

I have managed to convince myself the consumer side code is correct.

(Repeating some stuff that was said on irc:)

However, IMO having both sides allowed to update the read ptr is very
hazard-prone and confusing.

I suggested doing a conventional new protocol version (dm checks HV
feature, passes new version if available, both sides then speak new
protcol), where the new protocol either has each side doing % on each
write of its own pointer, or alternatively simply dropping the off by
one oddity in the ring size.

Jan replied:

07:10 <jbeulich> Diziet, andyhhp: Having each side canonicalize its pointer 
                 would break some of the comparisons of both pointers:
07:11 <jbeulich> Since readers would need to do modulo operations upon use, 
                 buffer full and buffer empty would become indistinguishable.

I'm not wholly convinced by this but I don't think I really want to
argue.  So I have applied the qemu-side patch to qemu-xen-traditional.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.