> -----Original Message----- > From: Sitsofe Wheeler > Sent: Thursday, August 28, 2014 20:49 PM > > > > The only issue seen on boot now is similar to > > > > https://lkml.org/lkml/2014/8/19/227 ... > > > I don't see this issue. Do you still see the issue for EVERY boot > > after you applied KY's always-use-page-allocation patch? I doubt that > > because in the log of the above link: > > I think it depends on if I do a UP or SMP boot. With > f1bd473f95e02bc382d4dae94d7f82e2a455e05d (post v3.17-rc2) with the V2 > BUG_ON > patch set coupled with the allocation change patch set a UP boot was able > to > run a small bunch of CPU and network stress tests without any issue. > However, > when doing an SMP boot the following happened: > <snip> > We can spin these off into a different thread if that would be helpful.
Hi Sitsofe, This seems a hv_netvsc specific issue(?) IMO it's better to open a new thread. However, I tried vcpus=1, 2 and 4 for 5 times respectively but couldn't reproduce the same issue(surely I used all of KY's patches, including the page-aligned-input-parameter-for-hypercall one) I used a workload of dd-ing and scp-ing big files. > > > > How come previous alignment efforts weren't working out? > > I'm not sure. > > If we trust the hypervisor, I would guess in hv_post_message() > > 1) We'd better add "aligned_msg->reserved = 0;" > > 2) Should we make sure "aligned_msg->payload_size % 8 == 0"? IMO > > aligned_msg->payload is an array of 8-byte. > > In that case why would payload_size not be a multiple of 8 - can it > change due to debug padding? If so wouldn't its start have had to be > misaligned? I found in some normal code path, e.g., vmbus_open() -> vmbus_post_msg() -> ... (here the payload size is sizeof(struct vmbus_channel_open_channel), i.e., 148, not a multiple of 8), the payload_size is not a multiple of 8. I don't think this causes the issue here, but I think we'd better double check this and see if there is a potential issue or not. Thanks, -- Dexuan _______________________________________________ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
