ARM: mm: implement no-highmem fast path in kmap_atomic_pfn()
PATCH 1/1 ARM: mm: implement no-highmem fast path in kmap_atomic_pfn()
Since CONFIG_HIGHMEM got enabled on ARMv5 Kirkwood, we have noticed a very significant drop in networking performance. The test were conducted on an OpenBlocks A7 board. Without this patch, the outgoing performance measured with iperf are:
- highmem OFF, TSO OFF 544 Mbit/s
- highmem OFF, TSO ON 942 Mbit/s
- highmem ON, TSO OFF 306 Mbit/s - highmem ON, TSO ON 246 Mbit/s
On this Kirkwood platform, the L2 cache is a Feroceon cache, and with this cache, all the range operations have to be done on virtual addresses and not physical addresses. Therefore, whenever CONFIG_HIGHMEM is enabled, the cache maintenance operations call kmap_atomic_pfn() and kunmap_atomic().
However, kmap_atomic_pfn() does not implement the same fast path for non-highmem pages as the one implemented in kmap_atomic(), and this is one of the reason for the performance drop. While this patch does not fully restore the performances, it clearly improves them a lot:
without patch with patch
- highmem ON, TSO OFF 306 Mbit/s 387 Mbit/s - highmem ON, TSO ON 246 Mbit/s 434 Mbit/s
We're still far from the !CONFIG_HIGHMEM performances, but it does improve a bit the situation.
Thanks a lot to Ezequiel Garcia and Gregory Clement for all the testing work around this topic.
--- arch/arm/mm/highmem.c | 3 +++ 1 file changed, 3 insertions(+)
arch/arm/mm/highmem.c +3 -0
@@ -127,8 +127,11 @@ void *kmap_atomic_pfn(unsigned long pfn) {unsigned long vaddr;int idx, type; +struct page *page =pfn_to_page (pfn);pagefault_disable (); +if (!PageHighMem (page)) +return page_address (page); type =kmap_atomic_idx_push (); idx = type +KM_TYPE_NR *smp_processor_id ();
Signature · 1 line
2.0.0
PATCH 1/1 ARM: mm: implement no-highmem fast path in kmap_atomic_pfn()
Quoted 43 lines · Thomas Petazzoni
Since
CONFIG_HIGHMEMgot enabled on ARMv5 Kirkwood, we have noticed a very significant drop in networking performance. The test were conducted on an OpenBlocks A7 board. Without this patch, the outgoing performance measured with iperf are:- highmem OFF, TSO OFF 544 Mbit/s
- highmem OFF, TSO ON 942 Mbit/s- highmem ON, TSO OFF 306 Mbit/s - highmem ON, TSO ON 246 Mbit/sOn this Kirkwood platform, the L2 cache is a Feroceon cache, and with this cache, all the range operations have to be done on virtual addresses and not physical addresses. Therefore, whenever
CONFIG_HIGHMEMis enabled, the cache maintenance operations callkmap_atomic_pfn()andkunmap_atomic().However,
kmap_atomic_pfn()does not implement the same fast path for non-highmem pages as the one implemented inkmap_atomic(), and this is one of the reason for the performance drop. While this patch does not fully restore the performances, it clearly improves them a lot:without patch with patch
- highmem ON, TSO OFF 306 Mbit/s 387 Mbit/s - highmem ON, TSO ON 246 Mbit/s 434 Mbit/sWe're still far from the !CONFIG_HIGHMEM performances, but it does improve a bit the situation.
Thanks a lot to Ezequiel Garcia and Gregory Clement for all the testing work around this topic.
Signed-off-by Thomas Petazzoni <[email protected]>--- arch/arm/mm/highmem.c | 3 +++ 1 file changed, 3 insertions(+)arch/arm/mm/highmem.c +0
-0@@ -127,8 +127,11 @@ void *kmap_atomic_pfn(unsigned long pfn) {
arch/arm/mm/highmem.c void *kmap_atomic_pfn(unsigned long pfn) · line 137
Quoted 10 lines
arch/arm/mm/highmem.c +3
-0unsigned long vaddr;int idx, type; +struct page *page =pfn_to_page (pfn);pagefault_disable (); +if (!PageHighMem (page)) +return page_address (page); type =kmap_atomic_idx_push (); idx = type +KM_TYPE_NR *smp_processor_id ();
What's the status of this one?
Signature · 3 lines
Ezequiel Garc?a, Free Electrons Embedded Linux, Kernel and Android Engineering http://free-electrons.com