]> git.kernelconcepts.de Git - karo-tx-linux.git/commit
clk-divider: make sure read-only dividers do not write to their register
authorHeiko Stuebner <heiko@sntech.de>
Thu, 21 Jan 2016 20:53:09 +0000 (21:53 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 11 May 2016 09:21:11 +0000 (11:21 +0200)
commitf0e92143b8e2e6fa1e854385667427011cfe1059
tree3ad99dd3ca9e2a7ee6416a22fdb8851fe0ddd68a
parentd65bf4e2407824aecba454c1b0bc7908a5113047
clk-divider: make sure read-only dividers do not write to their register

commit 50359819794b4a16ae35051cd80f2dab025f6019 upstream.

Commit e6d5e7d90be9 ("clk-divider: Fix READ_ONLY when divider > 1") removed
the special ops struct for read-only clocks and instead opted to handle
them inside the regular ops.

On the rk3368 this results in breakage as aclkm now gets set a value.
While it is the same divider value, the A53 core still doesn't like it,
which can result in the cpu ending up in a hang.
The reason being that "ACLKENMasserts one clock cycle before the rising
edge of ACLKM" and the clock should only be touched when STANDBYWFIL2
is asserted.

To fix this, reintroduce the read-only ops but do include the round_rate
callback. That way no writes that may be unsafe are done to the divider
register in any case.

The Rockchip use of the clk_divider_ops is adapted to this split again,
as is the nxp, lpc18xx-ccu driver that was included since the original
commit. On lpc18xx-ccu the divider seems to always be read-only
so only uses the new ops now.

Fixes: e6d5e7d90be9 ("clk-divider: Fix READ_ONLY when divider > 1")
Reported-by: Zhang Qing <zhangqing@rock-chips.com>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/clk/clk-divider.c
drivers/clk/nxp/clk-lpc18xx-ccu.c
drivers/clk/rockchip/clk.c
include/linux/clk-provider.h