diff --git a/debian/changelog b/debian/changelog index 80e9a1dbd..a1dee76d3 100644 --- a/debian/changelog +++ b/debian/changelog @@ -7,6 +7,10 @@ linux-2.6 (2.6.29-4) UNRELEASED; urgency=low * Extend erase timeout in M25P80 SPI Flash driver (Peter Horton). * Add driver for GMT G760A fan speed PWM controller chip. * [arm/orion5x] Enable SENSORS_G760A. + * Add patches from git.marvell.com: + - allow for alternative __copy_to_user/__clear_user implementations + - alternative copy_to_user/clear_user implementation copy_user + * [arm/orion5x, armel/kirkwood] Enable UACCESS_WITH_MEMCPY. [ dann frazier ] * [parisc] Fix macro expansion in atomic.h fixing PHONET compilation issue diff --git a/debian/config/arm/config.orion5x b/debian/config/arm/config.orion5x index f32eaa007..28beae437 100644 --- a/debian/config/arm/config.orion5x +++ b/debian/config/arm/config.orion5x @@ -609,3 +609,5 @@ CONFIG_X25=m CONFIG_SENSORS_G760A=m +CONFIG_UACCESS_WITH_MEMCPY=y + diff --git a/debian/config/armel/config.kirkwood b/debian/config/armel/config.kirkwood index ee85c12b0..5a3d80d99 100644 --- a/debian/config/armel/config.kirkwood +++ b/debian/config/armel/config.kirkwood @@ -631,3 +631,5 @@ CONFIG_SPI_ORION=y ## CONFIG_MTD_M25P80=y +CONFIG_UACCESS_WITH_MEMCPY=y + diff --git a/debian/patches/features/arm/allow-alternative-copy-user.patch b/debian/patches/features/arm/allow-alternative-copy-user.patch new file mode 100644 index 000000000..d40fa5ca3 --- /dev/null +++ b/debian/patches/features/arm/allow-alternative-copy-user.patch @@ -0,0 +1,63 @@ +From: Nicolas Pitre +Date: Mon, 9 Mar 2009 02:34:45 +0000 (-0400) +Subject: [ARM] allow for alternative __copy_to_user/__clear_user implementations +X-Git-Url: http://git.marvell.com/?p=orion.git;a=commitdiff_plain;h=3e4bd7ea3a16ec3a6f93f62a0d206df5b7f900e9 + +[ARM] allow for alternative __copy_to_user/__clear_user implementations + +This allows for optional alternative implementations of __copy_to_user +and __clear_user, with a possible runtime fallback to the standard +version when the alternative provides no gain over that standard +version. This is done by making the standard __copy_to_user into a weak +alias for the symbol __copy_to_user_std. Same thing for __clear_user. + +Those two functions are particularly good candidates to have alternative +implementations for, since they rely on the STRT instruction which has +lower performances than STM instructions on some CPU cores such as +the ARM1176 and Marvell Feroceon. + +Signed-off-by: Nicolas Pitre +--- + +diff --git a/arch/arm/include/asm/uaccess.h b/arch/arm/include/asm/uaccess.h +index 7897464..0da9bc9 100644 +--- a/arch/arm/include/asm/uaccess.h ++++ b/arch/arm/include/asm/uaccess.h +@@ -386,7 +386,9 @@ do { \ + #ifdef CONFIG_MMU + extern unsigned long __must_check __copy_from_user(void *to, const void __user *from, unsigned long n); + extern unsigned long __must_check __copy_to_user(void __user *to, const void *from, unsigned long n); ++extern unsigned long __must_check __copy_to_user_std(void __user *to, const void *from, unsigned long n); + extern unsigned long __must_check __clear_user(void __user *addr, unsigned long n); ++extern unsigned long __must_check __clear_user_std(void __user *addr, unsigned long n); + #else + #define __copy_from_user(to,from,n) (memcpy(to, (void __force *)from, n), 0) + #define __copy_to_user(to,from,n) (memcpy((void __force *)to, from, n), 0) +diff --git a/arch/arm/lib/clear_user.S b/arch/arm/lib/clear_user.S +index 4d6bc71..844f567 100644 +--- a/arch/arm/lib/clear_user.S ++++ b/arch/arm/lib/clear_user.S +@@ -18,7 +18,8 @@ + * : sz - number of bytes to clear + * Returns : number of bytes NOT cleared + */ +-ENTRY(__clear_user) ++ENTRY(__clear_user_std) ++WEAK(__clear_user) + stmfd sp!, {r1, lr} + mov r2, #0 + cmp r1, #4 +diff --git a/arch/arm/lib/copy_to_user.S b/arch/arm/lib/copy_to_user.S +index 22f968b..878820f 100644 +--- a/arch/arm/lib/copy_to_user.S ++++ b/arch/arm/lib/copy_to_user.S +@@ -86,7 +86,8 @@ + + .text + +-ENTRY(__copy_to_user) ++ENTRY(__copy_to_user_std) ++WEAK(__copy_to_user) + + #include "copy_template.S" + diff --git a/debian/patches/features/arm/alternative-copy-user.patch b/debian/patches/features/arm/alternative-copy-user.patch new file mode 100644 index 000000000..0dae4e1f8 --- /dev/null +++ b/debian/patches/features/arm/alternative-copy-user.patch @@ -0,0 +1,229 @@ +From: Lennert Buytenhek +Date: Mon, 9 Mar 2009 18:30:09 +0000 (-0400) +Subject: [ARM] alternative copy_to_user/clear_user implementation +X-Git-Url: http://git.marvell.com/?p=orion.git;a=commitdiff_plain;h=2d6c3846a588b82d1d6e13ef1afcbef8667ad272 + +[ARM] alternative copy_to_user/clear_user implementation + +This implements {copy_to,clear}_user() by faulting in the userland +pages and then using the regular kernel mem{cpy,set}() to copy the +data (while holding the page table lock). This is a win if the regular +mem{cpy,set}() implementations are faster than the user copy functions, +which is the case e.g. on Feroceon, where 8-word STMs (which memcpy() +uses under the right conditions) give significantly higher memory write +throughput than a sequence of individual 32bit stores. + +Here are numbers for page sized buffers on some Feroceon cores: + + - copy_to_user on Orion5x goes from 51 MB/s to 83 MB/s + - clear_user on Orion5x goes from 89MB/s to 314MB/s + - copy_to_user on Kirkwood goes from 240 MB/s to 356 MB/s + - clear_user on Kirkwood goes from 367 MB/s to 1108 MB/s + - copy_to_user on Disco-Duo goes from 248 MB/s to 398 MB/s + - clear_user on Disco-Duo goes from 328 MB/s to 1741 MB/s + +Because the setup cost is non negligible, this is worthwhile only if +the amount of data to copy is large enough. The operation falls back +to the standard implementation when the amount of data is below a certain +treshold. This treshold was determined empirically, however some targets +could benefit from a lower runtime determined value for optimal results +eventually. + +In the copy_from_user() case, this technique does not provide any +worthwhile performance gain due to the fact that any kind of read access +allocates the cache and subsequent 32bit loads are just as fast as the +equivalent 8-word LDM. + +[ Nicolas Pitre came with the idea, Lennert Buytenhek wrote the code. ] + +Tested-by: Martin Michlmayr +Signed-off-by: Lennert Buytenhek +Signed-off-by: Nicolas Pitre +--- + +diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig +index dbfdf87..0d1c72b 100644 +--- a/arch/arm/Kconfig ++++ b/arch/arm/Kconfig +@@ -983,6 +983,22 @@ config ALIGNMENT_TRAP + correct operation of some network protocols. With an IP-only + configuration it is safe to say N, otherwise say Y. + ++config UACCESS_WITH_MEMCPY ++ bool "Use kernel mem{cpy,set}() for {copy_to,clear}_user() (EXPERIMENTAL)" ++ depends on MMU && EXPERIMENTAL ++ default y if CPU_FEROCEON ++ help ++ Implement faster copy_to_user and clear_user methods for CPU ++ cores where a 8-word STM instruction give significantly higher ++ memory write throughput than a sequence of individual 32bit stores. ++ ++ A possible side effect is a slight increase in scheduling latency ++ between threads sharing the same address space if they invoke ++ such copy operations with large buffers. ++ ++ However, if the CPU data cache is using a write-allocate mode, ++ this option is unlikely to provide any performance gain. ++ + endmenu + + menu "Boot options" +diff --git a/arch/arm/lib/Makefile b/arch/arm/lib/Makefile +index 866f84a..030ba72 100644 +--- a/arch/arm/lib/Makefile ++++ b/arch/arm/lib/Makefile +@@ -29,6 +29,9 @@ else + endif + endif + ++# using lib_ here won't override already available weak symbols ++obj-$(CONFIG_UACCESS_WITH_MEMCPY) += uaccess_with_memcpy.o ++ + lib-$(CONFIG_MMU) += $(mmu-y) + + ifeq ($(CONFIG_CPU_32v3),y) +diff --git a/arch/arm/lib/uaccess_with_memcpy.c b/arch/arm/lib/uaccess_with_memcpy.c +new file mode 100644 +index 0000000..8bd9fc7 +--- /dev/null ++++ b/arch/arm/lib/uaccess_with_memcpy.c +@@ -0,0 +1,139 @@ ++/* ++ * linux/arch/arm/lib/uaccess_with_memcpy.c ++ * ++ * Copyright (C) 2009 Lennert Buytenhek ++ * Copyright (C) 2009 Marvell Semiconductor ++ * ++ * This program is free software; you can redistribute it and/or modify ++ * it under the terms of the GNU General Public License version 2 as ++ * published by the Free Software Foundation. ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include /* for in_atomic() */ ++#include ++#include ++ ++static int ++pin_page_for_write(const void __user *_addr, pte_t **ptep, spinlock_t **ptlp) ++{ ++ unsigned long addr = (unsigned long)_addr; ++ pgd_t *pgd; ++ pmd_t *pmd; ++ pte_t *pte; ++ spinlock_t *ptl; ++ ++ pgd = pgd_offset(current->mm, addr); ++ if (unlikely(pgd_none(*pgd) || pgd_bad(*pgd))) ++ return 0; ++ ++ pmd = pmd_offset(pgd, addr); ++ if (unlikely(pmd_none(*pmd) || pmd_bad(*pmd))) ++ return 0; ++ ++ pte = pte_offset_map_lock(current->mm, pmd, addr, &ptl); ++ if (unlikely(!pte_present(*pte) || !pte_young(*pte) || ++ !pte_write(*pte) || !pte_dirty(*pte))) { ++ pte_unmap_unlock(pte, ptl); ++ return 0; ++ } ++ ++ *ptep = pte; ++ *ptlp = ptl; ++ ++ return 1; ++} ++ ++unsigned long ++__copy_to_user(void __user *to, const void *from, unsigned long n) ++{ ++ int atomic; ++ ++ if (n < 1024) ++ return __copy_to_user_std(to, from, n); ++ ++ if (unlikely(segment_eq(get_fs(), KERNEL_DS))) { ++ memcpy((void *)to, from, n); ++ return 0; ++ } ++ ++ /* the mmap semaphore is taken only if not in an atomic context */ ++ atomic = in_atomic(); ++ ++ if (!atomic) ++ down_read(¤t->mm->mmap_sem); ++ while (n) { ++ pte_t *pte; ++ spinlock_t *ptl; ++ int tocopy; ++ ++ while (!pin_page_for_write(to, &pte, &ptl)) { ++ if (!atomic) ++ up_read(¤t->mm->mmap_sem); ++ if (__put_user(0, (char __user *)to)) ++ goto out; ++ if (!atomic) ++ down_read(¤t->mm->mmap_sem); ++ } ++ ++ tocopy = (~(unsigned long)to & ~PAGE_MASK) + 1; ++ if (tocopy > n) ++ tocopy = n; ++ ++ memcpy((void *)to, from, tocopy); ++ to += tocopy; ++ from += tocopy; ++ n -= tocopy; ++ ++ pte_unmap_unlock(pte, ptl); ++ } ++ if (!atomic) ++ up_read(¤t->mm->mmap_sem); ++ ++out: ++ return n; ++} ++ ++unsigned long __clear_user(void __user *addr, unsigned long n) ++{ ++ if (n < 256) ++ return __clear_user_std(addr, n); ++ ++ if (unlikely(segment_eq(get_fs(), KERNEL_DS))) { ++ memset((void *)addr, 0, n); ++ return 0; ++ } ++ ++ down_read(¤t->mm->mmap_sem); ++ while (n) { ++ pte_t *pte; ++ spinlock_t *ptl; ++ int tocopy; ++ ++ while (!pin_page_for_write(addr, &pte, &ptl)) { ++ up_read(¤t->mm->mmap_sem); ++ if (__put_user(0, (char __user *)addr)) ++ goto out; ++ down_read(¤t->mm->mmap_sem); ++ } ++ ++ tocopy = (~(unsigned long)addr & ~PAGE_MASK) + 1; ++ if (tocopy > n) ++ tocopy = n; ++ ++ memset((void *)addr, 0, tocopy); ++ addr += tocopy; ++ n -= tocopy; ++ ++ pte_unmap_unlock(pte, ptl); ++ } ++ up_read(¤t->mm->mmap_sem); ++ ++out: ++ return n; ++} diff --git a/debian/patches/series/4 b/debian/patches/series/4 index d0566ae8e..dab46e427 100644 --- a/debian/patches/series/4 +++ b/debian/patches/series/4 @@ -3,3 +3,5 @@ + bugfix/parisc/fix-macro-expansion-in-atomic.h.patch + bugfix/all/btrfs-fix-__ucmpdi2-compile-bug.patch + features/arm/g760a.patch ++ features/arm/allow-alternative-copy-user.patch ++ features/arm/alternative-copy-user.patch