Minki/linux
1
0
Fork 1
mirror of https://github.com/torvalds/linux.git synced 2024-12-04 18:13:04 +00:00
Code Issues Packages Projects Releases Wiki Activity
c8540870af
linux/arch/parisc/lib/lusercopy.S

363 lines
8.8 KiB
ArmAsm
Raw Normal View History

treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 150 Based on 1 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not write to the free software foundation inc 59 temple place suite 330 boston ma 02111 1307 usa extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 42 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Richard Fontana <rfontana@redhat.com> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190524100845.259718220@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2019-05-24 10:04:10 +00:00
/* SPDX-License-Identifier: GPL-2.0-or-later */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
/*
* User Space Access Routines
*
* Copyright (C) 2000-2002 Hewlett-Packard (John Marvin)
* Copyright (C) 2000 Richard Hirst <rhirst with parisc-linux.org>
* Copyright (C) 2001 Matthieu Delahaye <delahaym at esiee.fr>
* Copyright (C) 2003 Randolph Chung <tausq with parisc-linux.org>
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
* Copyright (C) 2017 Helge Deller <deller@gmx.de>
* Copyright (C) 2017 John David Anglin <dave.anglin@bell.net>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
*/
/*
* These routines still have plenty of room for optimization
* (word & doubleword load/store, dual issue, store hints, etc.).
*/
/*
* The following routines assume that space register 3 (sr3) contains
* the space id associated with the current users address space.
*/
Revert "parisc: fix trivial section name warnings" This reverts commit bd3bb8c15b9a80dbddfb7905b237a4a11a4725b4. Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
2008-05-22 18:36:31 +00:00
.text
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
#include <asm/assembly.h>
#include <asm/errno.h>
[PARISC] add ASM_EXCEPTIONTABLE_ENTRY() macro - this macro unifies the code to add exception table entries - additionally use ENTRY()/ENDPROC() at more places Signed-off-by: Helge Deller <deller@gmx.de> Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2007-01-28 13:52:57 +00:00
#include <linux/linkage.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
/*
* unsigned long lclear_user(void *to, unsigned long n)
*
* Returns 0 for success.
* otherwise, returns number of bytes not transferred.
*/
parisc: Add cfi_startproc and cfi_endproc to assembly code Add ENTRY_CFI() and ENDPROC_CFI() macros for dwarf debug info and convert assembly users to new macros. Signed-off-by: Helge Deller <deller@gmx.de>
2016-10-05 20:28:46 +00:00
ENTRY_CFI(lclear_user)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
comib,=,n 0,%r25,$lclu_done
$lclu_loop:
addib,<> -1,%r25,$lclu_loop
parisc: Implement __get/put_kernel_nofault() Remove CONFIG_SET_FS from parisc, so we need to add __get_kernel_nofault() and __put_kernel_nofault(), define HAVE_GET_KERNEL_NOFAULT and remove set_fs(), get_fs(), load_sr2(), thread_info->addr_limit, KERNEL_DS and USER_DS. The nice side-effect of this patch is that we now can directly access userspace via sr3 without the need to use a temporary sr2 which is either copied from sr3 or set to zero (for kernel space). Signed-off-by: Helge Deller <deller@gmx.de> Suggested-by: Arnd Bergmann <arnd@kernel.org>
2021-09-09 10:47:00 +00:00
1: stbs,ma %r0,1(%sr3,%r26)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
$lclu_done:
bv %r0(%r2)
copy %r25,%r28
parisc: Inline trivial exception code in lusercopy.S Fold trivial exception handling for lclear_user() and lstrnlen_user() into the main functions. Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-27 16:22:37 +00:00
2: b $lclu_done
ldo 1(%r25),%r25
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
parisc: provide macro to create exception table entries Provide a macro ASM_EXCEPTIONTABLE_ENTRY() to create exception table entries and convert all open-coded places to use that macro. This patch is a first step toward creating a exception table which only holds 32bit pointers even on a 64bit kernel. That way in my own kernel I was able to reduce the in-kernel exception table from 44kB to 22kB. Signed-off-by: Helge Deller <deller@gmx.de>
2013-10-13 19:11:30 +00:00
ASM_EXCEPTIONTABLE_ENTRY(1b,2b)
parisc: Inline trivial exception code in lusercopy.S Fold trivial exception handling for lclear_user() and lstrnlen_user() into the main functions. Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-27 16:22:37 +00:00
ENDPROC_CFI(lclear_user)
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
/*
* unsigned long pa_memcpy(void *dstp, const void *srcp, unsigned long len)
*
* Inputs:
* - sr1 already contains space of source region
* - sr2 already contains space of destination region
*
* Returns:
* - number of bytes that could not be copied.
* On success, this will be zero.
*
* This code is based on a C-implementation of a copy routine written by
* Randolph Chung, which in turn was derived from the glibc.
*
* Several strategies are tried to try to get the best performance for various
* conditions. In the optimal case, we copy by loops that copy 32- or 16-bytes
* at a time using general registers. Unaligned copies are handled either by
* aligning the destination and then using shift-and-write method, or in a few
* cases by falling back to a byte-at-a-time copy.
*
* Testing with various alignments and buffer sizes shows that this code is
* often >10x faster than a simple byte-at-a-time copy, even for strangely
* aligned operands. It is interesting to note that the glibc version of memcpy
* (written in C) is actually quite fast already. This routine is able to beat
* it by 30-40% for aligned copies because of the loop unrolling, but in some
* cases the glibc version is still slightly faster. This lends more
* credibility that gcc can generate very good code as long as we are careful.
*
* Possible optimizations:
* - add cache prefetching
* - try not to use the post-increment address modifiers; they may create
* additional interlocks. Assumption is that those were only efficient on old
* machines (pre PA8000 processors)
*/
dst = arg0
src = arg1
len = arg2
end = arg3
t1 = r19
t2 = r20
t3 = r21
t4 = r22
srcspc = sr1
dstspc = sr2
t0 = r1
a1 = t1
a2 = t2
a3 = t3
a0 = t4
save_src = ret0
save_dst = ret1
save_len = r31
ENTRY_CFI(pa_memcpy)
/* Last destination address */
add dst,len,end
/* short copy with less than 16 bytes? */
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
cmpib,COND(>>=),n 15,len,.Lbyte_loop
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
/* same alignment? */
xor src,dst,t0
extru t0,31,2,t1
cmpib,<>,n 0,t1,.Lunaligned_copy
#ifdef CONFIG_64BIT
/* only do 64-bit copies if we can get aligned. */
extru t0,31,3,t1
cmpib,<>,n 0,t1,.Lalign_loop32
/* loop until we are 64-bit aligned */
.Lalign_loop64:
extru dst,31,3,t1
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
cmpib,=,n 0,t1,.Lcopy_loop_16_start
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
20: ldb,ma 1(srcspc,src),t1
21: stb,ma t1,1(dstspc,dst)
b .Lalign_loop64
ldo -1(len),len
ASM_EXCEPTIONTABLE_ENTRY(20b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(21b,.Lcopy_done)
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
.Lcopy_loop_16_start:
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
ldi 31,t0
.Lcopy_loop_16:
cmpb,COND(>>=),n t0,len,.Lword_loop
10: ldd 0(srcspc,src),t1
11: ldd 8(srcspc,src),t2
ldo 16(src),src
12: std,ma t1,8(dstspc,dst)
13: std,ma t2,8(dstspc,dst)
14: ldd 0(srcspc,src),t1
15: ldd 8(srcspc,src),t2
ldo 16(src),src
16: std,ma t1,8(dstspc,dst)
17: std,ma t2,8(dstspc,dst)
ASM_EXCEPTIONTABLE_ENTRY(10b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(11b,.Lcopy16_fault)
ASM_EXCEPTIONTABLE_ENTRY(12b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(13b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(14b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(15b,.Lcopy16_fault)
ASM_EXCEPTIONTABLE_ENTRY(16b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(17b,.Lcopy_done)
b .Lcopy_loop_16
ldo -32(len),len
.Lword_loop:
cmpib,COND(>>=),n 3,len,.Lbyte_loop
20: ldw,ma 4(srcspc,src),t1
21: stw,ma t1,4(dstspc,dst)
b .Lword_loop
ldo -4(len),len
ASM_EXCEPTIONTABLE_ENTRY(20b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(21b,.Lcopy_done)
#endif /* CONFIG_64BIT */
/* loop until we are 32-bit aligned */
.Lalign_loop32:
extru dst,31,2,t1
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
cmpib,=,n 0,t1,.Lcopy_loop_8
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
20: ldb,ma 1(srcspc,src),t1
21: stb,ma t1,1(dstspc,dst)
b .Lalign_loop32
ldo -1(len),len
ASM_EXCEPTIONTABLE_ENTRY(20b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(21b,.Lcopy_done)
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
.Lcopy_loop_8:
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
cmpib,COND(>>=),n 15,len,.Lbyte_loop
10: ldw 0(srcspc,src),t1
11: ldw 4(srcspc,src),t2
12: stw,ma t1,4(dstspc,dst)
13: stw,ma t2,4(dstspc,dst)
14: ldw 8(srcspc,src),t1
15: ldw 12(srcspc,src),t2
ldo 16(src),src
16: stw,ma t1,4(dstspc,dst)
17: stw,ma t2,4(dstspc,dst)
ASM_EXCEPTIONTABLE_ENTRY(10b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(11b,.Lcopy8_fault)
ASM_EXCEPTIONTABLE_ENTRY(12b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(13b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(14b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(15b,.Lcopy8_fault)
ASM_EXCEPTIONTABLE_ENTRY(16b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(17b,.Lcopy_done)
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
b .Lcopy_loop_8
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
ldo -16(len),len
.Lbyte_loop:
cmpclr,COND(<>) len,%r0,%r0
b,n .Lcopy_done
20: ldb 0(srcspc,src),t1
ldo 1(src),src
21: stb,ma t1,1(dstspc,dst)
b .Lbyte_loop
ldo -1(len),len
ASM_EXCEPTIONTABLE_ENTRY(20b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(21b,.Lcopy_done)
.Lcopy_done:
bv %r0(%r2)
sub end,dst,ret0
/* src and dst are not aligned the same way. */
/* need to go the hard way */
.Lunaligned_copy:
/* align until dst is 32bit-word-aligned */
extru dst,31,2,t1
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
cmpib,=,n 0,t1,.Lcopy_dstaligned
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
20: ldb 0(srcspc,src),t1
ldo 1(src),src
21: stb,ma t1,1(dstspc,dst)
b .Lunaligned_copy
ldo -1(len),len
ASM_EXCEPTIONTABLE_ENTRY(20b,.Lcopy_done)
ASM_EXCEPTIONTABLE_ENTRY(21b,.Lcopy_done)
.Lcopy_dstaligned:
/* store src, dst and len in safe place */
copy src,save_src
copy dst,save_dst
copy len,save_len
/* len now needs give number of words to copy */
SHRREG len,2,len
/*
* Copy from a not-aligned src to an aligned dst using shifts.
* Handles 4 words per loop.
*/
depw,z src,28,2,t0
subi 32,t0,t0
mtsar t0
extru len,31,2,t0
cmpib,= 2,t0,.Lcase2
/* Make src aligned by rounding it down. */
depi 0,31,2,src
cmpiclr,<> 3,t0,%r0
b,n .Lcase3
cmpiclr,<> 1,t0,%r0
b,n .Lcase1
.Lcase0:
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
cmpb,COND(=) %r0,len,.Lcda_finish
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
nop
1: ldw,ma 4(srcspc,src), a3
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
1: ldw,ma 4(srcspc,src), a0
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
b,n .Ldo3
.Lcase1:
1: ldw,ma 4(srcspc,src), a2
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
1: ldw,ma 4(srcspc,src), a3
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
ldo -1(len),len
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
cmpb,COND(=),n %r0,len,.Ldo0
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
.Ldo4:
1: ldw,ma 4(srcspc,src), a0
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
shrpw a2, a3, %sar, t0
1: stw,ma t0, 4(dstspc,dst)
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcopy_done)
.Ldo3:
1: ldw,ma 4(srcspc,src), a1
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
shrpw a3, a0, %sar, t0
1: stw,ma t0, 4(dstspc,dst)
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcopy_done)
.Ldo2:
1: ldw,ma 4(srcspc,src), a2
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
shrpw a0, a1, %sar, t0
1: stw,ma t0, 4(dstspc,dst)
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcopy_done)
.Ldo1:
1: ldw,ma 4(srcspc,src), a3
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
shrpw a1, a2, %sar, t0
1: stw,ma t0, 4(dstspc,dst)
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcopy_done)
ldo -4(len),len
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
cmpb,COND(<>) %r0,len,.Ldo4
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
nop
.Ldo0:
shrpw a2, a3, %sar, t0
1: stw,ma t0, 4(dstspc,dst)
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcopy_done)
.Lcda_rdfault:
.Lcda_finish:
/* calculate new src, dst and len and jump to byte-copy loop */
sub dst,save_dst,t0
add save_src,t0,src
b .Lbyte_loop
sub save_len,t0,len
.Lcase3:
1: ldw,ma 4(srcspc,src), a0
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
1: ldw,ma 4(srcspc,src), a1
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
b .Ldo2
ldo 1(len),len
.Lcase2:
1: ldw,ma 4(srcspc,src), a1
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
1: ldw,ma 4(srcspc,src), a2
ASM_EXCEPTIONTABLE_ENTRY(1b,.Lcda_rdfault)
b .Ldo1
ldo 2(len),len
/* fault exception fixup handlers: */
#ifdef CONFIG_64BIT
.Lcopy16_fault:
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
b .Lcopy_done
10: std,ma t1,8(dstspc,dst)
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
ASM_EXCEPTIONTABLE_ENTRY(10b,.Lcopy_done)
#endif
.Lcopy8_fault:
parisc: fix bugs in pa_memcpy The patch 554bfeceb8a22d448cd986fc9efce25e833278a1 ("parisc: Fix access fault handling in pa_memcpy()") reimplements the pa_memcpy function. Unfortunatelly, it makes the kernel unbootable. The crash happens in the function ide_complete_cmd where memcpy is called with the same source and destination address. This patch fixes a few bugs in pa_memcpy: * When jumping to .Lcopy_loop_16 for the first time, don't skip the instruction "ldi 31,t0" (this bug made the kernel unbootable) * Use the COND macro when comparing length, so that the comparison is 64-bit (a theoretical issue, in case the length is greater than 0xffffffff) * Don't use the COND macro after the "extru" instruction (the PA-RISC specification says that the upper 32-bits of extru result are undefined, although they are set to zero in practice) * Fix exception addresses in .Lcopy16_fault and .Lcopy8_fault * Rename .Lcopy_loop_4 to .Lcopy_loop_8 (so that it is consistent with .Lcopy8_fault) Cc: <stable@vger.kernel.org> # v4.9+ Fixes: 554bfeceb8a2 ("parisc: Fix access fault handling in pa_memcpy()") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Helge Deller <deller@gmx.de>
2017-04-14 18:15:20 +00:00
b .Lcopy_done
10: stw,ma t1,4(dstspc,dst)
parisc: Fix access fault handling in pa_memcpy() pa_memcpy() is the major memcpy implementation in the parisc kernel which is used to do any kind of userspace/kernel memory copies. Al Viro noticed various bugs in the implementation of pa_mempcy(), most notably that in case of faults it may report back to have copied more bytes than it actually did. Fixing those bugs is quite hard in the C-implementation, because the compiler is messing around with the registers and we are not guaranteed that specific variables are always in the same processor registers. This makes proper fault handling complicated. This patch implements pa_memcpy() in assembler. That way we have correct fault handling and adding a 64-bit copy routine was quite easy. Runtime tested with 32- and 64bit kernels. Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: <stable@vger.kernel.org> # v4.9+ Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
2017-03-29 19:41:05 +00:00
ASM_EXCEPTIONTABLE_ENTRY(10b,.Lcopy_done)
ENDPROC_CFI(pa_memcpy)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
.end
Reference in New Issue Copy Permalink