From mboxrd@z Thu Jan 1 00:00:00 1970 From: Linus Torvalds Date: Sat, 29 Mar 2008 02:52:06 +0000 Subject: Re: [PATCH 2/4] set_restore_sigmask TIF_SIGPENDING Message-Id: List-Id: References: <20080329001230.D013726FA1D@magilla.localdomain> <20080329001341.7F93826FA1D@magilla.localdomain> <20080329022408.0DD4726FA1D@magilla.localdomain> In-Reply-To: <20080329022408.0DD4726FA1D@magilla.localdomain> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Roland McGrath Cc: Andrew Morton , Martin Schwidefsky , linux-s390@vger.kernel.org, tony.luck@intel.com, linux-ia64@vger.kernel.org, linux-arch@vger.kernel.org, linux-kernel@vger.kernel.org On Fri, 28 Mar 2008, Roland McGrath wrote: > > It could be a PF_* too, I suppose. There aren't too many of those > bits free, but it would have the advantage of being a place for an > arch that doesn't store any TS_* bits anywhere. Yeah, I guess PF_ would be a bit more regular. Maybe we should even try to avoid the use of TS_ in x86, and turn it into PF_. There are probably bad historical reasons for the duplication of capabilities. > Since acting on the flag is in arch signal code anyway, it makes some > sense to let the arch define how it gets that to happen. I'll send > some follow-on patches that change the conditionals to use #ifdef > HAVE_SET_RESTORE_SIGMASK. Let's see if it matters first. No reason to add another arch-specific thing if nobody can even measure this thing, and from a quick look it seems like every RESTORE_SIGMASK user is basically an error path for a system call. Those few extra cycles really won't be noticeable, we almost certainly have better things we could use our energy on. So never mind. I think your series is fine, and my TS_ idea doesn't really look like it's worth it (and using PF_ sounds a bit more palatable since we could do it with existing infrastructure, but a quick grep shows that there's more users of test_thread_flag(TIF_RESTORE_SIGMASK) than I would have expected (and the *testing* is equally cheap for atomic and thread- synchronous fields, so that's not a performance issue). Linus