From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from pentafluge.infradead.org (pentafluge.infradead.org [213.146.154.40]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTP id 31885DDEF3 for ; Mon, 18 Jun 2007 19:24:58 +1000 (EST) Subject: Re: [PATCH] Wire up sys_sync_file_range() on powerpc From: David Woodhouse To: Paul Mackerras In-Reply-To: <18038.6658.854660.4233@cargo.ozlabs.ibm.com> References: <1181992681.25228.683.camel@pmac.infradead.org> <1182018044.2808.75.camel@pmac.infradead.org> <18038.6658.854660.4233@cargo.ozlabs.ibm.com> Content-Type: text/plain Date: Mon, 18 Jun 2007 17:25:11 +0800 Message-Id: <1182158711.2799.78.camel@shinybook.infradead.org> Mime-Version: 1.0 Cc: akpm@osdl.org, linux-arch@vger.kernel.org, linuxppc-dev@ozlabs.org, torvalds@osdl.org, drepper@redhat.com, rmk@arm.linux.org.uk List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Mon, 2007-06-18 at 15:37 +1000, Paul Mackerras wrote: > David Woodhouse writes: > > > Pants. > > Ah. Is that a naughty word in the UK? If so I'll have to remember > that when I go to kernel summit. :) I think the most important thing you have to remember about 'pants' while in the UK is that you don't wear them on the outside. > > It doesn't work because the 'flags' argument ends up in r9, and > > we can only use r3-r8 for syscall arguments. We'll need to do it the > > same way as ARM does, with the flags as the second argument. > > Yes. We'll need a wrapper for both 32-bit and for 32-on-64. > > > Or is it? Can we ditch sys_sync_file_range now and implement a new > > sys_sync_file_range2 with the two 32-bit arguments first? > > Would be nice... It might look something like this... I can't test it sensibly till I get home next week; anyone else want to poke at it? (Note that glibc's syscall() doesn't get 64-bit arguments like this right; you'll want to implement do_sync_file_range() in assembly if you're using Andrew's original test program.) .section ".text" .globl do_sync_file_range .type do_sync_file_range, @function do_sync_file_range: stwu 1,-16(1) mflr 0 stw 31,12(1) stw 0,20(1) li 0,308 sc mfcr 0 andis. 9,0,4096 mr 31,3 beq 0,.L6 bl __errno_location li 0,-1 stw 31,0(3) .L6: mr 3,0 lwz 0,20(1) lwz 31,12(1) addi 1,1,16 mtlr 0 blr .size do_sync_file_range,.-do_sync_file_range --- [PATCH] Add sys_sync_file_range2() system call ... with more sensible ABI, for those architectures which hated sys_sync_file_range() for wasting argument slots by misaligning 64-bit values. Convert ARM to use it, wire it up on PowerPC, and prevent the missing syscall checker from bitching about sync_file_range if you have sync_file_range2 instead. Signed-off-by: David Woodhouse diff --git a/arch/arm/kernel/calls.S b/arch/arm/kernel/calls.S index 19326d7..a98d0c9 100644 --- a/arch/arm/kernel/calls.S +++ b/arch/arm/kernel/calls.S @@ -350,7 +350,7 @@ CALL(sys_set_robust_list) CALL(sys_get_robust_list) /* 340 */ CALL(sys_splice) - CALL(sys_arm_sync_file_range) + CALL(sys_sync_file_range2) CALL(sys_tee) CALL(sys_vmsplice) CALL(sys_move_pages) diff --git a/arch/arm/kernel/sys_arm.c b/arch/arm/kernel/sys_arm.c index 1ca2d51..4d25e49 100644 --- a/arch/arm/kernel/sys_arm.c +++ b/arch/arm/kernel/sys_arm.c @@ -328,16 +328,3 @@ asmlinkage long sys_arm_fadvise64_64(int fd, int advice, { return sys_fadvise64_64(fd, offset, len, advice); } - -/* - * Yet more syscall fsckage - we can't fit sys_sync_file_range's - * arguments into the available registers with EABI. So, let's - * create an ARM specific syscall for this which has _sane_ - * arguments. (This incidentally also has an ABI-independent - * argument layout.) - */ -asmlinkage long sys_arm_sync_file_range(int fd, unsigned int flags, - loff_t offset, loff_t nbytes) -{ - return sys_sync_file_range(fd, offset, nbytes, flags); -} diff --git a/arch/powerpc/kernel/sys_ppc32.c b/arch/powerpc/kernel/sys_ppc32.c index 047246a..b42cbf1 100644 --- a/arch/powerpc/kernel/sys_ppc32.c +++ b/arch/powerpc/kernel/sys_ppc32.c @@ -810,3 +810,12 @@ asmlinkage long compat_sys_request_key(const char __user *_type, return sys_request_key(_type, _description, _callout_info, destringid); } +asmlinkage long compat_sys_sync_file_range2(int fd, unsigned int flags, + unsigned offset_hi, unsigned offset_lo, + unsigned nbytes_hi, unsigned nbytes_lo) +{ + loff_t offset = ((loff_t)offset_hi << 32) | offset_lo; + loff_t nbytes = ((loff_t)nbytes_hi << 32) | nbytes_lo; + + return sys_sync_file_range(fd, offset, nbytes, flags); +} diff --git a/fs/sync.c b/fs/sync.c index 2f97576..7cd005e 100644 --- a/fs/sync.c +++ b/fs/sync.c @@ -236,6 +236,14 @@ out: return ret; } +/* It would be nice if people remember that not all the world's an i386 + when they introduce new system calls */ +asmlinkage long sys_sync_file_range2(int fd, unsigned int flags, + loff_t offset, loff_t nbytes) +{ + return sys_sync_file_range(fd, offset, nbytes, flags); +} + /* * `endbyte' is inclusive */ diff --git a/include/asm-arm/unistd.h b/include/asm-arm/unistd.h index 250d7f1..bfdbebe 100644 --- a/include/asm-arm/unistd.h +++ b/include/asm-arm/unistd.h @@ -367,6 +367,7 @@ #define __NR_get_robust_list (__NR_SYSCALL_BASE+339) #define __NR_splice (__NR_SYSCALL_BASE+340) #define __NR_arm_sync_file_range (__NR_SYSCALL_BASE+341) +#define __NR_sync_file_range2 __NR_arm_sync_file_range #define __NR_tee (__NR_SYSCALL_BASE+342) #define __NR_vmsplice (__NR_SYSCALL_BASE+343) #define __NR_move_pages (__NR_SYSCALL_BASE+344) diff --git a/include/asm-powerpc/systbl.h b/include/asm-powerpc/systbl.h index 700ca59..1cc3f9c 100644 --- a/include/asm-powerpc/systbl.h +++ b/include/asm-powerpc/systbl.h @@ -311,3 +311,4 @@ COMPAT_SYS_SPU(utimensat) COMPAT_SYS_SPU(signalfd) COMPAT_SYS_SPU(timerfd) SYSCALL_SPU(eventfd) +COMPAT_SYS_SPU(sync_file_range2) diff --git a/include/asm-powerpc/unistd.h b/include/asm-powerpc/unistd.h index e3c28dc..f71c606 100644 --- a/include/asm-powerpc/unistd.h +++ b/include/asm-powerpc/unistd.h @@ -330,10 +330,11 @@ #define __NR_signalfd 305 #define __NR_timerfd 306 #define __NR_eventfd 307 +#define __NR_sync_file_range2 308 #ifdef __KERNEL__ -#define __NR_syscalls 308 +#define __NR_syscalls 309 #define __NR__exit __NR_exit #define NR_syscalls __NR_syscalls diff --git a/scripts/checksyscalls.sh b/scripts/checksyscalls.sh index f98171f..0dcc01c 100755 --- a/scripts/checksyscalls.sh +++ b/scripts/checksyscalls.sh @@ -99,6 +99,11 @@ cat << EOF #define __IGNORE_setfsuid32 #define __IGNORE_setfsgid32 +/* sync_file_range had a stupid ABI. Allow sync_file_range2 instead */ +#ifdef __NR_sync_file_range2 +#define __IGNORE_sync_file_range +#endif + /* Unmerged syscalls for AFS, STREAMS, etc. */ #define __IGNORE_afs_syscall #define __IGNORE_getpmsg -- dwmw2