From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mitchel Humpherys Subject: Re: [PATCH RESEND v8] iopoll: Introduce memory-mapped IO polling macros Date: Mon, 24 Nov 2014 17:21:09 -0800 Message-ID: References: <1416860092-11620-1-git-send-email-mitchelh@codeaurora.org> <94D0CD8314A33A4D9D801C0FE68B4029593C5E1F@G9W0745.americas.hpqcorp.net> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <94D0CD8314A33A4D9D801C0FE68B4029593C5E1F-W1gbDvblbosSZAcGdq5asR6epYMZPwEe5NbjCUgZEJk@public.gmane.org> (Robert Elliott's message of "Tue, 25 Nov 2014 00:53:19 +0000") List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: iommu-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org Errors-To: iommu-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org To: "Elliott, Robert (Server Storage)" Cc: "Scales, Webb" , Arnd Bergmann , Will Deacon , "linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" , "Don Brace (PMC)" , "iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org" , Thierry Reding , Matt Wagantall , Andrew Morton , "linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org" List-Id: iommu@lists.linux-foundation.org On Mon, Nov 24 2014 at 04:53:19 PM, "Elliott, Robert (Server Storage)" wrote: >> -----Original Message----- >> From: linux-kernel-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org [mailto:linux-kernel- >> owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org] On Behalf Of Mitchel Humpherys >> Sent: Monday, 24 November, 2014 2:15 PM > ... >> From: Matt Wagantall >> >> It is sometimes necessary to poll a memory-mapped register until its value >> satisfies some condition. Introduce a family of convenience macros that do >> this. Tight-looping, sleeping, and timing out can all be accomplished >> using these macros. >> > ... >> +#define readx_poll_timeout(op, addr, val, cond, sleep_us, timeout_us) \ >> +({ \ >> + ktime_t timeout = ktime_add_us(ktime_get(), timeout_us); \ >> + might_sleep_if(sleep_us); \ >> + for (;;) { \ >> + (val) = op(addr); \ >> + if (cond) \ >> + break; \ >> + if (timeout_us && ktime_compare(ktime_get(), timeout) > 0) { >> \ >> + (val) = op(addr); \ >> + break; \ >> + } \ >> + if (sleep_us) \ >> + usleep_range((sleep_us >> 2) + 1, sleep_us); \ > > The hpsa SCSI driver used to use usleep_range in a loop like > that, but we found that it caused scheduler problems during > boots because it uses TASK_UNINTERRUPTIBLE: > [ 9.260668] [sched_delayed] sched: RT throttling activated > > msleep() worked much better. Hmm, maybe you were just sleeping for too long? According to Documentation/timers/timers-howto.txt, usleep_range is what should be used for non-atomic sleeps in the range [10us, 20ms]. Plus we need microsecond granularity anyways, so msleep wouldn't cut it. If there are any potential users of these macros that would want to sleep for more than 20ms I guess we could add a special case here to use msleep when sleep_us exceeds 20,000 or so. -Mitch -- Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project