diff options
| author | Laurent Vivier <laurent@vivier.eu> | 2022-04-06 22:15:20 +0200 |
|---|---|---|
| committer | Geert Uytterhoeven <geert@linux-m68k.org> | 2022-04-11 11:48:01 +0200 |
| commit | 2e2ac4a3327479f7e2744cdd88a5c823f2057bad (patch) | |
| tree | b35193673d8e7fdac3f13e2c007a005c71ab664b /drivers/clocksource/timer-goldfish.c | |
| parent | 3123109284176b1532874591f7c81f3837bbdc17 (diff) | |
| download | linux-2e2ac4a3327479f7e2744cdd88a5c823f2057bad.tar.bz2 | |
tty: goldfish: Introduce gf_ioread32()/gf_iowrite32()
The goldfish TTY device was clearly defined as having little-endian
registers, but the switch to __raw_{read,write}l(() broke its driver
when running on big-endian kernels (if anyone ever tried this).
The m68k qemu implementation got this wrong, and assumed native-endian
registers. While this is a bug in qemu, it is probably impossible to
fix that since there is no way of knowing which other operating systems
have started relying on that bug over the years.
Hence revert commit da31de35cd2f ("tty: goldfish: use
__raw_writel()/__raw_readl()", and define gf_ioread32()/gf_iowrite32()
to be able to use accessors defined by the architecture.
Cc: stable@vger.kernel.org # v5.11+
Fixes: da31de35cd2fb78f ("tty: goldfish: use __raw_writel()/__raw_readl()")
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Link: https://lore.kernel.org/r/20220406201523.243733-2-laurent@vivier.eu
[geert: Add rationale based on Arnd's comments]
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Diffstat (limited to 'drivers/clocksource/timer-goldfish.c')
0 files changed, 0 insertions, 0 deletions