ARM: no longer force unbuffered DMA for realview

Commit 42c4dafe80 ("ARM: 6202/1: Do not ARM_DMA_MEM_BUFFERABLE
on RealView boards with L210/L220") changed the generic setting for
ARM_DMA_MEM_BUFFERABLE to be disabled on any Realview kernel that includes
support for any of the ARM11 variations. Doing this was required to
allow doing DMA without a lockup in the l2x0 cache controller on the
Realview platform.

Unfortunately, in a kernel that also contains support for any ARMv7
based machine, the same change makes it impossible to do DMA on ARMv7,
which gets in the way of enabling multiplatform support on Realview.

As confirmed by Catalin Marinas and Linus Walleij, the current
code for Realview that we have in the kernel does not actually
perform any DMA, and this is unlikely to change in the future.
Therefore we can revert 42c4dafe80 without introducing regressions,
but we must never start using DMA on this platform in the future.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
This commit is contained in:
Arnd Bergmann 2015-11-25 17:32:17 +01:00 committed by Linus Walleij
parent 3c30a4a357
commit 38541bf485
1 changed files with 0 additions and 2 deletions

View File

@ -1005,8 +1005,6 @@ config ARM_L1_CACHE_SHIFT
config ARM_DMA_MEM_BUFFERABLE
bool "Use non-cacheable memory for DMA" if (CPU_V6 || CPU_V6K) && !CPU_V7
depends on !(MACH_REALVIEW_PB1176 || REALVIEW_EB_ARM11MP || \
MACH_REALVIEW_PB11MP)
default y if CPU_V6 || CPU_V6K || CPU_V7
help
Historically, the kernel has used strongly ordered mappings to