Re: [PATCH 1/1] ppc64: fix missing to check all bits of _TIF_USER_WORK_MASK in preempt
From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Date: 2011-12-13 05:55:14
From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Date: 2011-12-13 05:55:14
On Tue, 2011-12-13 at 13:01 +0800, tiejun.chen wrote:
Tiejun Chen wrote:quoted
In entry_64.S version of ret_from_except_lite, you'll notice that in the !preempt case, after we've checked MSR_PR we test for any TIF flag in _TIF_USER_WORK_MASK to decide whether to go to do_work or not. However, in the preempt case, we do a convoluted trick to test SIGPENDING only if PR was set and always test NEED_RESCHED ... but we forget to test any other bit of _TIF_USER_WORK_MASK !!! So that means that with preempt, we completely fail to test for things like single step, syscall tracing, etc... This should be fixed as the following path: - Test PR. If set, go to test_work_user, else continue. - In test_work_user, always test for _TIF_USER_WORK_MASK to decide to go to do_work, maybe call it do_user_work - In test_work_kernel, test for _TIF_KERNEL_WORK_MASK which is set to our new flag along with NEED_RESCHED if preempt is enabled and branch to do_kernel_work.Ben, Any comment for this?
Sorry, I didn't get to review that one yet (nor reply to your newer responses), I have very sore eyes and basically had to get off the computer. Hopefully I'll be better tomorrow. Cheers, Ben.