]> www.infradead.org Git - users/willy/pagecache.git/commit
arm64: mm: Fix zone_dma_limit calculation
authorYang Shi <yang@os.amperecomputing.com>
Mon, 25 Nov 2024 17:16:50 +0000 (09:16 -0800)
committerCatalin Marinas <catalin.marinas@arm.com>
Tue, 3 Dec 2024 17:22:39 +0000 (17:22 +0000)
commit56a708742a8bf127eb66798bfc9c9516c61f9930
tree8e9bff327d6af4f893a88820a79097552c9fd28e
parent49ccf2c3cafb5774a4562d61294afcf492bed487
arm64: mm: Fix zone_dma_limit calculation

Commit ba0fb44aed47 ("dma-mapping: replace zone_dma_bits by
zone_dma_limit") and subsequent patches changed how zone_dma_limit is
calculated to allow a reduced ZONE_DMA even when RAM starts above 4GB.
Commit 122c234ef4e1 ("arm64: mm: keep low RAM dma zone") further fixed
this to ensure ZONE_DMA remains below U32_MAX if RAM starts below 4GB,
especially on platforms that do not have IORT or DT description of the
device DMA ranges. While zone boundaries calculation was fixed by the
latter commit, zone_dma_limit, used to determine the GFP_DMA flag in the
core code, was not updated. This results in excessive use of GFP_DMA and
unnecessary ZONE_DMA allocations on some platforms.

Update zone_dma_limit to match the actual upper bound of ZONE_DMA.

Fixes: ba0fb44aed47 ("dma-mapping: replace zone_dma_bits by zone_dma_limit")
Cc: <stable@vger.kernel.org> # 6.12.x
Reported-by: Yutang Jiang <jiangyutang@os.amperecomputing.com>
Tested-by: Yutang Jiang <jiangyutang@os.amperecomputing.com>
Signed-off-by: Yang Shi <yang@os.amperecomputing.com>
Link: https://lore.kernel.org/r/20241125171650.77424-1-yang@os.amperecomputing.com
[catalin.marinas@arm.com: some tweaking of the commit log]
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
arch/arm64/mm/init.c