[PATCH v2 0/2] mark driver as non-removable

[Russell King - ARM Linux]

On Wed, Dec 21, 2016 at 04:03:38PM +0100, Alexander Stein wrote:
> Changes in v2;
> * Instead of adding a remove function which is unused otherwise, mark the driver
>   as non-remoable
> 
> Using DEBUG_TEST_DRIVER_REMOVE every driver gets probed, removed and probed
> again. This breaks drivers without removal function, e.g. arm perf
> resulting in the following dump:
> ------------[ cut here ]------------
> WARNING: CPU: 1 PID: 1 at fs/sysfs/dir.c:31 sysfs_warn_dup+0x6c/0x7c
> sysfs: cannot create duplicate filename '/devices/armv7_cortex_a7'
> Modules linked in:
> CPU: 1 PID: 1 Comm: swapper/0 Not tainted 4.9.0+ #212
> Hardware name: Freescale LS1021A
> Backtrace:
> [<8020c044>] (dump_backtrace) from [<8020c2f0>] (show_stack+0x18/0x1c)
>  r7:00000009 r6:60000013 r5:80c1776c r4:00000000
> [<8020c2d8>] (show_stack) from [<8046ead8>] (dump_stack+0x94/0xa8)
> [<8046ea44>] (dump_stack) from [<8021ecd0>] (__warn+0xec/0x104)
>  r7:00000009 r6:8092efc8 r5:00000000 r4:bf04bd80
> [<8021ebe4>] (__warn) from [<8021ed28>] (warn_slowpath_fmt+0x40/0x48)
>  r9:80a32848 r8:00000000 r7:00000000 r6:bf0ab780 r5:8091d590 r4:bf0df000
> [<8021ecec>] (warn_slowpath_fmt) from [<8036d53c>] (sysfs_warn_dup+0x6c/0x7c)
>  r3:bf0df000 r2:8092ef94
> [<8036d4d0>] (sysfs_warn_dup) from [<8036d628>] (sysfs_create_dir_ns+0x8c/0x9c)
>  r6:bf0ab780 r5:bf20ee08 r4:ffffffef
> [<8036d59c>] (sysfs_create_dir_ns) from [<80471860>] (kobject_add_internal+0xa8/0x34c)
>  r6:bf0aa198 r5:00000000 r4:bf20ee08
> [<804717b8>] (kobject_add_internal) from [<80471b54>] (kobject_add+0x50/0x94)
>  r7:00000000 r6:00000000 r5:00000000 r4:bf20ee08
> [<80471b08>] (kobject_add) from [<80524590>] (device_add+0xe4/0x590)
>  r3:00000000 r2:00000000
>  r6:bf20ee00 r5:00000000 r4:bf20ee08
> [<805244ac>] (device_add) from [<802a38a8>] (pmu_dev_alloc+0x88/0xd8)
>  r10:00000091 r9:80a32848 r8:00000000 r7:80a32840 r6:80c0ed3c r5:00000000
>  r4:bf1fe800
> [<802a3820>] (pmu_dev_alloc) from [<80a0cd20>] (perf_event_sysfs_init+0x5c/0xb4)
>  r7:80a32840 r6:00000000 r5:bf1fe800 r4:80c0ede0
> [<80a0ccc4>] (perf_event_sysfs_init) from [<80201778>] (do_one_initcall+0x48/0x178)
>  r6:80c45000 r5:80a0ccc4 r4:00000006
> [<80201730>] (do_one_initcall) from [<80a00ecc>] (kernel_init_freeable+0x168/0x20c)
>  r8:80a00610 r7:80a32840 r6:80c45000 r5:80c45000 r4:00000006
> [<80a00d64>] (kernel_init_freeable) from [<806febcc>] (kernel_init+0x10/0x11c)
>  r10:00000000 r9:00000000 r8:00000000 r7:00000000 r6:00000000 r5:806febbc
>  r4:00000000
> [<806febbc>] (kernel_init) from [<80208388>] (ret_from_fork+0x14/0x2c)
>  r5:806febbc r4:00000000
> ---[ end trace 9d251d389382804f ]---

Please consider putting some of the above in patch 2's description, so
the reason for the patch doesn't get lost.  If you want to reduce the
commit message, you could consider cutting the registers from the
backtrace (which are only useful when doing in-depth debugging, not for
illustrating the callpath.)

I'd like to see acks on both of these before I take patch 2, but I'm not
expecting that to happen until the new year.

Thanks.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.