kernel_optimize_test/arch/arm/mm
Russell King 2be23c475a ARM: Ensure PTE modifications via dma_alloc_coherent are visible
Dave Hylands reports:
| We've observed a problem with dma_alloc_writecombine when the system
| is under heavy load (heavy bus traffic).  We've managed to reduce the
| problem to the following snippet, which is run from a kthread in a
| continuous loop:
|
|   void *virtAddr;
|   dma_addr_t physAddr;
|   unsigned int numBytes = 256;
|
|   for (;;) {
|       virtAddr = dma_alloc_writecombine(NULL,
|             numBytes, &physAddr, GFP_KERNEL);
|       if (virtAddr == NULL) {
|          printk(KERN_ERR "Running out of memory\n");
|          break;
|       }
|
|       /* access DMA memory allocated */
|       tmp = virtAddr;
|       *tmp = 0x77;
|
|       /* free DMA memory */
|       dma_free_writecombine(NULL,
|             numBytes, virtAddr, physAddr);
|
|         ...sleep here...
|     }
|
| By itself, the code will run forever with no issues. However, as we
| increase our bus traffic (typically using DMA) then the *tmp = 0x77
| line will eventually cause a page fault. If we add a small delay (a
| few microseconds) before the *tmp = 0x77, then we don't see a page
| fault, even under heavy load.

A dsb() is required after modifying the PTE entries to ensure that they
will always be visible.  Add this dsb().

Reported-by: Dave Hylands <dhylands@gmail.com>
Tested-by: Dave Hylands <dhylands@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2010-09-08 16:27:56 +01:00
..
abort-ev4.S
abort-ev4t.S
abort-ev5t.S
abort-ev5tj.S
abort-ev6.S
abort-ev7.S
abort-lv4t.S
abort-macro.S
abort-nommu.S
alignment.c ARM: 6268/1: ARMv6K and ARMv7 use fault statuses 3 and 6 as Access Flag fault 2010-07-27 10:48:41 +01:00
cache-fa.S
cache-feroceon-l2.c
cache-l2x0.c ARM: 6272/1: Convert L2x0 to use the IO relaxed operations 2010-07-29 14:04:36 +01:00
cache-tauros2.c
cache-v3.S
cache-v4.S
cache-v4wb.S
cache-v4wt.S
cache-v6.S
cache-v7.S
cache-xsc3l2.c
context.c
copypage-fa.c
copypage-feroceon.c
copypage-v3.c
copypage-v4mc.c
copypage-v4wb.c
copypage-v4wt.c
copypage-v6.c
copypage-xsc3.c
copypage-xscale.c
dma-mapping.c ARM: Ensure PTE modifications via dma_alloc_coherent are visible 2010-09-08 16:27:56 +01:00
extable.c
fault-armv.c
fault.c ARM: 6268/1: ARMv6K and ARMv7 use fault statuses 3 and 6 as Access Flag fault 2010-07-27 10:48:41 +01:00
fault.h
flush.c
highmem.c kmap_atomic: make kunmap_atomic() harder to misuse 2010-08-09 20:44:54 -07:00
init.c Merge branch 'devel-stable' into devel 2010-07-31 14:20:16 +01:00
iomap.c
ioremap.c ARM: use generic ioremap_page_range() 2010-07-27 10:43:47 +01:00
Kconfig ARM: 6344/1: Mark CPU_32v6K as depended on CPU_V7 2010-09-02 15:32:13 +01:00
Makefile
mm.h ARM: Convert platform reservations to use LMB rather than bootmem 2010-07-27 08:48:23 +01:00
mmap.c
mmu.c Merge branch 'misc' into devel 2010-07-31 14:20:02 +01:00
nommu.c ARM: Convert platform reservations to use LMB rather than bootmem 2010-07-27 08:48:23 +01:00
pabort-legacy.S
pabort-v6.S
pabort-v7.S
pgd.c
proc-arm6_7.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm7tdmi.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm9tdmi.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm720.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm740.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm920.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm922.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm925.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm926.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm940.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm946.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm1020.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm1020e.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm1022.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-arm1026.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-fa526.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-feroceon.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-macros.S
proc-mohawk.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-sa110.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-sa1100.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-syms.c
proc-v6.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-v7.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-xsc3.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
proc-xscale.S ARM: Factor out common code from cpu_proc_fin() 2010-07-27 10:48:42 +01:00
tlb-fa.S
tlb-v3.S
tlb-v4.S
tlb-v4wb.S
tlb-v4wbi.S
tlb-v6.S
tlb-v7.S
vmregion.c ARM: DMA coherent allocator: align remapped addresses 2010-07-27 10:43:48 +01:00
vmregion.h ARM: DMA coherent allocator: align remapped addresses 2010-07-27 10:43:48 +01:00