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

Re: [Xen-devel] [PATCH] Refactor ioreq server for better performance



[snip]


Thanks, Paul.
Well, I agree the former approach would be simpler. But I still doubt
if this is more reasonable. :)
IIUC, one of the reasons for struct domain to have a rangeset list(and
a spinlock - rangesets_lock), is because there are iomem_caps and
irq_caps for each domain. These 2 rangeset members of struct domain are
platform independent.
However, struct rb_rangeset is only supposed to be used in ioreq
server, which is only for x86 hvm cases. Adding a rb_rangeset list
member(similarly, if so, a rb_rangesets_lock is also required) in
struct domain maybe useless for hardware domain and for platforms other
than x86.

Fair enough.

So, I'd like to register a new debug key, to dump the ioreq server
informations, just like the keys to dump iommu p2m table or the irq
mappings. With a new debug key, we do not need to add a spinlock for
rb_rangeset in struct domain, the one in ioreq server would be enough.
Does this sound reasonable?


That would be ok with me, but I'm not sure about claiming a whole debug key for 
this. Is there any other one that you could piggy-back on? If not, then maybe 
just make it part of the 'q' output.

Thanks, my new implementation uses the 'q' debug key. Will send out the
new version later. :)

Yu

   Paul


[snip]



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel





_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel


 


Rackspace

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