On Sat, Jul 20, 2013 at 11:39 PM, Johannes Berg <johannes@xxxxxxxxxxxxxxxx> wrote: > On Fri, 2013-07-19 at 15:13 -0700, Luis R. Rodriguez wrote: >> From: Johannes Berg <johannes.berg@xxxxxxxxx> >> >> Since kernel version 3.3, workqueue names could be >> sprintf'ed instead of just pointing to a name. This >> wasn't used a lot so never needed to be backported, >> but now it's used everywhere. Backport this API. >> >> Signed-off-by: Johannes Berg <johannes.berg@xxxxxxxxx> >> [mcgrof: add WQ_HIGHPRI and WQ_MEM_RECLAIM defines] > > Were they needed somehow? Thye shouldn't have been, since users of this > don't usually use the flags directly? There's not used but at compile time the flags arguments were kept, we can surely nuke them if we #ifdef a call for older kernels. > If they _are_ needed, then we probably need to do something different? > My patch only backported "regular" and "ordered" workqueue, not the > other flags, and I suspect the other flags can't actually be backported. Indeed they can't, we just ignore them for older kernels, in fact perhaps just using create_singlethread_workqueue(name) would suffice for older kernels. >> +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,36) >> +#define WQ_HIGHPRI 0 >> +#define WQ_MEM_RECLAIM 0 >> #endif > > Ah, here. I think that would cause problems? Not necessarily -- these are simply not usable / backportable on older kernels. In part of my synch up of patches (see d7608058) I nuked patches/collateral-evolutions/network/46-use_other_workqueue/ which used to backport alloc_workqueue() with create_singlethread_workqueue() given older kernels do not support WQ_HIGHPRI and WQ_MEM_RECLAIM. In your port you make use of __create_workqueue_key() but that's available in older kernels as well, its just the flags are kept in code, but they are not used per se. >> +static DEFINE_SPINLOCK(wq_name_lock); >> +static struct list_head wq_name_list; > > FWIW, my patch was [RFC] for a reason, I hadn't tested it... Now that I > did, I can tell you that this must be > > static LIST_HEAD(wq_name_list); > > to get the list properly initialised. I'll add that. Sorry for jumping the gun here. Luis -- To unsubscribe from this list: send the line "unsubscribe backports" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
