]> code.ossystems Code Review - openembedded-core.git/commit
linux-yocto/5.4: qemuppc32: reduce serial shutdown issues
authorBruce Ashfield <bruce.ashfield@gmail.com>
Fri, 7 May 2021 16:22:57 +0000 (12:22 -0400)
committerRichard Purdie <richard.purdie@linuxfoundation.org>
Sun, 9 May 2021 22:30:10 +0000 (23:30 +0100)
commit42355cb73049ee7a4af0f539a2a5b7d4ee1abc65
treed8613c4d9bb89c4d285c479877c8ed3276ed5592
parent17f18d0aa3a8588717b6b5ae2a2b9c773fbf7863
linux-yocto/5.4: qemuppc32: reduce serial shutdown issues

Integrating the following commit(s) to linux-yocto/5.4:

    qemuppc32: reduce serial issues seen on shutdown

    Richard reported:

    We've been seeing a lot of the qemuppc shutdown issue and I decided to
    look into it. The really worrying thing looking at the logs locally is the
    serial ports are showing irq issues and becoming disabled as nobody would
    handle them.

    Errors like:

       [    9.194886] irq 36: nobody cared (try booting with the "irqpoll" option)
       [    9.198712] CPU: 0 PID: 127 Comm: bootlogd Not tainted
       [    9.202283] Call Trace:
       [    9.205611] [d1005f00] [c00a0da8] __report_bad_irq+0x50/0x138 (unreliable)
       [    9.209347] [d1005f30] [c00a0cc0] note_interrupt+0x324/0x378
       [    9.212855] [d1005f70] [c009d138] handle_irq_event+0xe8/0x104
       [    9.216353] [d1005fa0] [c00a1d9c] handle_fasteoi_irq+0xc0/0x29c
       [    9.219960] [d1005fc0] [c009b798] generic_handle_irq+0x40/0x5c
       [    9.223496] [d1005fd0] [c00075d0] __do_irq+0x58/0x188
       [    9.226948] [d1005ff0] [c0010040] call_do_irq+0x20/0x38
       [    9.230391] [d29eda60] [c0007788] do_IRQ+0x88/0xfc
       [    9.233860] [d29eda90] [c0016454] ret_from_except+0x0/0x14
       [    9.237288] --- interrupt: 501 at __setup_irq+0x3c4/0x838
       [    9.237288]     LR = __setup_irq+0x790/0x838
       [    9.244155] [d29edb88] [c009f0a4] request_threaded_irq+0x114/0x1c8
       [    9.247672] [d29edbb8] [c07a5a18] pmz_startup+0x17c/0x32c
       [    9.251203] [d29edbd8] [c07a1140] uart_port_startup+0x184/0x2f8
       [    9.254651] [d29edc08] [c07a1974] uart_port_activate+0x78/0xf4
       [    9.258141] [d29edc28] [c07839f8] tty_port_open+0xd4/0x170
       [    9.261579] [d29edc58] [c079db74] uart_open+0x2c/0x48
       [    9.265116] [d29edc68] [c077a288] tty_open+0x168/0x640
       [    9.268574] [d29edcd8] [c0280be8] chrdev_open+0x138/0x2a4
       [    9.272123] [d29edd18] [c027421c] do_dentry_open+0x228/0x410
       [    9.275643] [d29edd48] [c028e9f4] path_openat+0xb04/0xf28
       [    9.279184] [d29eddd8] [c02917e4] do_filp_open+0x120/0x164
       [    9.282535] [d29ede98] [c0276238] do_sys_openat2+0xd8/0x19c
       [    9.285790] [d29edee8] [c0276574] sys_openat+0x88/0xdc
       [    9.289096] [d29edf38] [c00160d8] ret_from_syscall+0x0/0x34
       [    9.292620] --- interrupt: c01 at 0xfec3738
       [    9.292620]     LR = 0xfec36e0
       [    9.299035] handlers:
       [    9.302312] [<7f7f7da8>] pmz_interrupt
       [    9.305541] Disabling IRQ #36

    (and the irqpoll option does not help)

    This is problematic as the shutdown test uses the serial interface to
    shut down the system. If the serial interface fails to login or run the command,
    game over for the test.

    CONFIG_SERIAL_PMACZILOG_CONSOLE complicates that handling, but doesn't provide
    any output or capabilities that we need. So we disable it here, and
    reduce the chances of issues during shutdown.

Signed-off-by: Bruce Ashfield <bruce.ashfield@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
meta/recipes-kernel/linux/linux-yocto-rt_5.4.bb
meta/recipes-kernel/linux/linux-yocto-tiny_5.4.bb
meta/recipes-kernel/linux/linux-yocto_5.4.bb