tty: goldfish: Introduce gf_ioread32()/gf_iowrite32()
authorLaurent Vivier <laurent@vivier.eu>
Wed, 6 Apr 2022 20:15:20 +0000 (22:15 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 9 Jun 2022 08:23:20 +0000 (10:23 +0200)
commit0b011b408f349e76f7903a477870051f94c7e119
treebd666186c16b16321b3ac7833223fd84bde4ad53
parentb3485d2b09e946da83e530112e7eea0da997b96a
tty: goldfish: Introduce gf_ioread32()/gf_iowrite32()

commit 2e2ac4a3327479f7e2744cdd88a5c823f2057bad upstream.

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>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/tty/goldfish.c
include/linux/goldfish.h