]> git.kernelconcepts.de Git - karo-tx-linux.git/commit
mwifiex: don't enable/disable IRQ 0 during suspend/resume
authorBrian Norris <briannorris@chromium.org>
Fri, 10 Feb 2017 21:55:25 +0000 (13:55 -0800)
committerKalle Valo <kvalo@codeaurora.org>
Wed, 15 Feb 2017 09:14:10 +0000 (11:14 +0200)
commit2447e2cad75239ae407c0f98acf12511354208c5
tree4efc194e404b39ba6d55d552025a50ed9a407ad9
parent3a6282045b228f6de6abced74b793c71178fb948
mwifiex: don't enable/disable IRQ 0 during suspend/resume

If we don't have an out-of-band wakeup IRQ configured through DT (as
most platforms don't), then we fall out of this function with
'irq_wakeup == 0'. Other code (e.g., mwifiex_disable_wake() and
mwifiex_enable_wake()) treats 'irq_wakeup >= 0' as a valid IRQ, and so
we end up calling {enable,disable}_irq() on IRQ 0.

That seems bad, so let's not do that.

Same problem as fixed in this patch:

https://patchwork.kernel.org/patch/9531693/
[PATCH v2 2/3] btmrvl: set irq_bt to -1 when failed to parse it

with the difference that:
(a) this one is actually a regression and
(b) this affects both device tree and non-device-tree systems

While fixing the regression, also drop the verbosity on the parse
failure, so we don't see this when a DT node is present but doesn't have
an interrupt property (this is perfectly legal):

[   21.999000] mwifiex_pcie 0000:01:00.0: fail to parse irq_wakeup from device tree

Fixes: 853402a00823 ("mwifiex: Enable WoWLAN for both sdio and pcie")
Signed-off-by: Brian Norris <briannorris@chromium.org>
Acked-by: Rajat Jain <rajatja@google.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
drivers/net/wireless/marvell/mwifiex/main.c