On 12/07/2015 08:33 PM, Pratyush Anand wrote: > On 07/12/2015:02:51:16 PM, peter@xxxxxxxxxxxxxxxxxx wrote: >>> On Dec 7, 2015, at 7:44 AM, Pratyush Anand <panand@xxxxxxxxxx> wrote: >>>> On 07/12/2015:06:54:13 AM, peter@xxxxxxxxxxxxxxxxxx wrote: >>>>> On Dec 7, 2015, at 3:50 AM, Pratyush Anand <panand@xxxxxxxxxx> wrote: >>>>> >>>>> Hi tty experts, >>>>> >>>>> One of of our application (related to kdump) is failing because of a kernel >>>>> change in tty layer. It fails to start "Kdump Vmcore Save Service" because it is >>>>> not able to redirect all standard input/output messages to console (exec &> >>>>> /dev/console). >>>> >>>> What device is /dev/console? >>> >>> Its amba-pl011. >>> >>>> What device is stdin/stdout? >>> >>> stdin is not being used in this case. stdout is again serial console >>> (amba-pl011). However, if I replace "exec &> /dev/console" with "echo test > >>> /dev/console", then also it gets same error. I'm confused; if stdout is the same tty as /dev/console, why the redirection? >> Hmm. What's odd is that the race you're observing is only possible >> when the tty is in final close which isn't true if stdin and stdout >> and /dev/console all have the same underlying tty device. >> >>>> When do you observe this failure? >>> >>> During a early systemd service called after init. >>> >>> I observe this during kdump operation on seattle platform. When primary kernel >>> crashes, kdump loads a secondary kernel and initramfs which copies vmcore(memory >>> dump of failed kernel) to the disk for further analysis. It initializes some >>> systemd services before starting vmcore copy. One of the early systemd service, >>> sets up (mainly) STDOUT to /dev/console, so that copy operation message can be >>> viewed. And this operation fails. >> >> I'm traveling right now but will have a chance to spend time on this Wed. > > Thanks :-) > >> >> I'm sure patches I already posted on list will address this, but >> I'd like to make sure the cause is a race on final close, and not >> some other oversight. > > I applied "[PATCH v2] Replace tty->closing" series on top of v4.4-rc3, but I am > still able to reproduce it. That series is not applicable to this problem. The series, "Rework tty_reopen()", will likely fix this. Note that series also requires a much earlier series, "tty core printk cleanup". Regards, Peter Hurley -- To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
