Messages in this thread | | | Date | Wed, 12 Mar 2014 18:49:31 +0000 | From | Zoltan Kiss <> | Subject | Re: [PATCH net-next v7 0/9] xen-netback: TX grant mapping with SKBTX_DEV_ZEROCOPY instead of copy |
| |
On 12/03/14 15:40, Ian Campbell wrote: > On Sat, 2014-03-08 at 18:57 -0500, David Miller wrote: >> From: Zoltan Kiss <zoltan.kiss@citrix.com> >> Date: Sat, 8 Mar 2014 14:37:50 +0000 >> >>> Maybe you mixed up mine with that? But that's also not eligible to be >>> applied yet. >> >> I can always revert the series if there are major objections. > > Zoltan -- does this patch series suffer from/expose the confusion > regarding RING_HAS_UNCONSUMED_REQUESTS which we are discussing > separately on xen-devel? If the answer is yes then I think this series > should be reverted for the time being because there seems to be some > fairly fundamental questions about the semantics of that macro. I haven't seen it causing any issue during my testing, although it went through several XenRT nighlies. That topic ("RING_HAS_UNCONSUMED_REQUESTS oddness" on xen-devel) came from theoretical grounds. One outcome of it is that we should move that napi_schedule from the callback to the end of the dealloc thread to be on the safe side. I can post a short patch for that.
> > If the answer is no then I will endeavour to review this version of the > series ASAP (hopefully tomorrow) and determine if I have any other major > objections which would warrant a revert. > > Ian. >
| |