Messages in this thread | | | Date | Thu, 29 Sep 2011 12:37:15 -0700 | From | Andrew Morton <> | Subject | Re: [159/244] ipc/mqueue.c: fix mq_open() return value |
| |
On Thu, 29 Sep 2011 12:08:55 -0700 Greg KH <gregkh@suse.de> wrote:
> > And in fact, upon further reflection, I think maybe that particular > > test could use being split into two distinct tests. One for wrapping > > the byte counter, which would return -ENOMEM, and one for exceeding > > RLIMIT_MSGQUEUE which would return -EPERM (not sure if that's right, I > > would have to poke around elsewhere, but it seems a better response > > when you are violating a ulimit than nomem to me anyway). > > Ok, care to get the patch into Linus's tree and then I can take it into > stable?
Doug, the thing to do here is to rework your recent mqueue patchset. Prepare a minimal, critical-stuff series of bugfix patches against current Linus mainline which is also applicable to -stable. We can merge that into 3.1 or, more likely, into 3.2-rc1/3.1.x.
Then, later, we can merge up the less critical parts of that patchset.
otoh, the only not-applicable-to-stable part of that patchset appears to be "[1/5] ipc/mqueue: cleanup definition names and locations" and it's fairly small. So we could perhaps just merge all five into 3.2-rc1, with a -stable backport.
Doing that backport would require that we first backport the buggy patches (ie: this one), which is a bit weird. Perhaps it would be better for you to prepare a reworked patch series for 3.0.x after that five-patch series hits mainline.
| |