summaryrefslogtreecommitdiffstats
path: root/arch/xtensa/lib
diff options
context:
space:
mode:
Diffstat (limited to 'arch/xtensa/lib')
-rw-r--r--arch/xtensa/lib/checksum.S74
-rw-r--r--arch/xtensa/lib/memcopy.S81
-rw-r--r--arch/xtensa/lib/memset.S45
-rw-r--r--arch/xtensa/lib/pci-auto.c45
-rw-r--r--arch/xtensa/lib/strncpy_user.S60
-rw-r--r--arch/xtensa/lib/strnlen_user.S28
-rw-r--r--arch/xtensa/lib/usercopy.S138
7 files changed, 185 insertions, 286 deletions
diff --git a/arch/xtensa/lib/checksum.S b/arch/xtensa/lib/checksum.S
index 4eb573d2720e..528fe0dd9339 100644
--- a/arch/xtensa/lib/checksum.S
+++ b/arch/xtensa/lib/checksum.S
@@ -14,9 +14,10 @@
* 2 of the License, or (at your option) any later version.
*/
-#include <asm/errno.h>
+#include <linux/errno.h>
#include <linux/linkage.h>
#include <variant/core.h>
+#include <asm/asmmacro.h>
/*
* computes a partial checksum, e.g. for TCP/UDP fragments
@@ -175,23 +176,8 @@ ENDPROC(csum_partial)
/*
* Copy from ds while checksumming, otherwise like csum_partial
- *
- * The macros SRC and DST specify the type of access for the instruction.
- * thus we can call a custom exception handler for each access type.
*/
-#define SRC(y...) \
- 9999: y; \
- .section __ex_table, "a"; \
- .long 9999b, 6001f ; \
- .previous
-
-#define DST(y...) \
- 9999: y; \
- .section __ex_table, "a"; \
- .long 9999b, 6002f ; \
- .previous
-
/*
unsigned int csum_partial_copy_generic (const char *src, char *dst, int len,
int sum, int *src_err_ptr, int *dst_err_ptr)
@@ -244,28 +230,28 @@ ENTRY(csum_partial_copy_generic)
add a10, a10, a2 /* a10 = end of last 32-byte src chunk */
.Loop5:
#endif
-SRC( l32i a9, a2, 0 )
-SRC( l32i a8, a2, 4 )
-DST( s32i a9, a3, 0 )
-DST( s32i a8, a3, 4 )
+EX(10f) l32i a9, a2, 0
+EX(10f) l32i a8, a2, 4
+EX(11f) s32i a9, a3, 0
+EX(11f) s32i a8, a3, 4
ONES_ADD(a5, a9)
ONES_ADD(a5, a8)
-SRC( l32i a9, a2, 8 )
-SRC( l32i a8, a2, 12 )
-DST( s32i a9, a3, 8 )
-DST( s32i a8, a3, 12 )
+EX(10f) l32i a9, a2, 8
+EX(10f) l32i a8, a2, 12
+EX(11f) s32i a9, a3, 8
+EX(11f) s32i a8, a3, 12
ONES_ADD(a5, a9)
ONES_ADD(a5, a8)
-SRC( l32i a9, a2, 16 )
-SRC( l32i a8, a2, 20 )
-DST( s32i a9, a3, 16 )
-DST( s32i a8, a3, 20 )
+EX(10f) l32i a9, a2, 16
+EX(10f) l32i a8, a2, 20
+EX(11f) s32i a9, a3, 16
+EX(11f) s32i a8, a3, 20
ONES_ADD(a5, a9)
ONES_ADD(a5, a8)
-SRC( l32i a9, a2, 24 )
-SRC( l32i a8, a2, 28 )
-DST( s32i a9, a3, 24 )
-DST( s32i a8, a3, 28 )
+EX(10f) l32i a9, a2, 24
+EX(10f) l32i a8, a2, 28
+EX(11f) s32i a9, a3, 24
+EX(11f) s32i a8, a3, 28
ONES_ADD(a5, a9)
ONES_ADD(a5, a8)
addi a2, a2, 32
@@ -284,8 +270,8 @@ DST( s32i a8, a3, 28 )
add a10, a10, a2 /* a10 = end of last 4-byte src chunk */
.Loop6:
#endif
-SRC( l32i a9, a2, 0 )
-DST( s32i a9, a3, 0 )
+EX(10f) l32i a9, a2, 0
+EX(11f) s32i a9, a3, 0
ONES_ADD(a5, a9)
addi a2, a2, 4
addi a3, a3, 4
@@ -315,8 +301,8 @@ DST( s32i a9, a3, 0 )
add a10, a10, a2 /* a10 = end of last 2-byte src chunk */
.Loop7:
#endif
-SRC( l16ui a9, a2, 0 )
-DST( s16i a9, a3, 0 )
+EX(10f) l16ui a9, a2, 0
+EX(11f) s16i a9, a3, 0
ONES_ADD(a5, a9)
addi a2, a2, 2
addi a3, a3, 2
@@ -326,8 +312,8 @@ DST( s16i a9, a3, 0 )
4:
/* This section processes a possible trailing odd byte. */
_bbci.l a4, 0, 8f /* 1-byte chunk */
-SRC( l8ui a9, a2, 0 )
-DST( s8i a9, a3, 0 )
+EX(10f) l8ui a9, a2, 0
+EX(11f) s8i a9, a3, 0
#ifdef __XTENSA_EB__
slli a9, a9, 8 /* shift byte to bits 8..15 */
#endif
@@ -350,10 +336,10 @@ DST( s8i a9, a3, 0 )
add a10, a10, a2 /* a10 = end of last odd-aligned, 2-byte src chunk */
.Loop8:
#endif
-SRC( l8ui a9, a2, 0 )
-SRC( l8ui a8, a2, 1 )
-DST( s8i a9, a3, 0 )
-DST( s8i a8, a3, 1 )
+EX(10f) l8ui a9, a2, 0
+EX(10f) l8ui a8, a2, 1
+EX(11f) s8i a9, a3, 0
+EX(11f) s8i a8, a3, 1
#ifdef __XTENSA_EB__
slli a9, a9, 8 /* combine into a single 16-bit value */
#else /* for checksum computation */
@@ -381,7 +367,7 @@ ENDPROC(csum_partial_copy_generic)
a12 = original dst for exception handling
*/
-6001:
+10:
_movi a2, -EFAULT
s32i a2, a6, 0 /* src_err_ptr */
@@ -403,7 +389,7 @@ ENDPROC(csum_partial_copy_generic)
2:
retw
-6002:
+11:
movi a2, -EFAULT
s32i a2, a7, 0 /* dst_err_ptr */
movi a2, 0
diff --git a/arch/xtensa/lib/memcopy.S b/arch/xtensa/lib/memcopy.S
index b1c219acabe7..c0f6981719d6 100644
--- a/arch/xtensa/lib/memcopy.S
+++ b/arch/xtensa/lib/memcopy.S
@@ -9,23 +9,9 @@
* Copyright (C) 2002 - 2012 Tensilica Inc.
*/
+#include <linux/linkage.h>
#include <variant/core.h>
-
- .macro src_b r, w0, w1
-#ifdef __XTENSA_EB__
- src \r, \w0, \w1
-#else
- src \r, \w1, \w0
-#endif
- .endm
-
- .macro ssa8 r
-#ifdef __XTENSA_EB__
- ssa8b \r
-#else
- ssa8l \r
-#endif
- .endm
+#include <asm/asmmacro.h>
/*
* void *memcpy(void *dst, const void *src, size_t len);
@@ -123,10 +109,8 @@
addi a5, a5, 2
j .Ldstaligned # dst is now aligned, return to main algorithm
- .align 4
- .global memcpy
- .type memcpy,@function
-memcpy:
+ENTRY(__memcpy)
+WEAK(memcpy)
entry sp, 16 # minimal stack frame
# a2/ dst, a3/ src, a4/ len
@@ -209,7 +193,7 @@ memcpy:
.Lsrcunaligned:
_beqz a4, .Ldone # avoid loading anything for zero-length copies
# copy 16 bytes per iteration for word-aligned dst and unaligned src
- ssa8 a3 # set shift amount from byte offset
+ __ssa8 a3 # set shift amount from byte offset
/* set to 1 when running on ISS (simulator) with the
lint or ferret client, or 0 to save a few cycles */
@@ -229,16 +213,16 @@ memcpy:
.Loop2:
l32i a7, a3, 4
l32i a8, a3, 8
- src_b a6, a6, a7
+ __src_b a6, a6, a7
s32i a6, a5, 0
l32i a9, a3, 12
- src_b a7, a7, a8
+ __src_b a7, a7, a8
s32i a7, a5, 4
l32i a6, a3, 16
- src_b a8, a8, a9
+ __src_b a8, a8, a9
s32i a8, a5, 8
addi a3, a3, 16
- src_b a9, a9, a6
+ __src_b a9, a9, a6
s32i a9, a5, 12
addi a5, a5, 16
#if !XCHAL_HAVE_LOOPS
@@ -249,10 +233,10 @@ memcpy:
# copy 8 bytes
l32i a7, a3, 4
l32i a8, a3, 8
- src_b a6, a6, a7
+ __src_b a6, a6, a7
s32i a6, a5, 0
addi a3, a3, 8
- src_b a7, a7, a8
+ __src_b a7, a7, a8
s32i a7, a5, 4
addi a5, a5, 8
mov a6, a8
@@ -261,7 +245,7 @@ memcpy:
# copy 4 bytes
l32i a7, a3, 4
addi a3, a3, 4
- src_b a6, a6, a7
+ __src_b a6, a6, a7
s32i a6, a5, 0
addi a5, a5, 4
mov a6, a7
@@ -288,14 +272,14 @@ memcpy:
s8i a6, a5, 0
retw
+ENDPROC(__memcpy)
/*
* void bcopy(const void *src, void *dest, size_t n);
*/
- .align 4
- .global bcopy
- .type bcopy,@function
-bcopy:
+
+ENTRY(bcopy)
+
entry sp, 16 # minimal stack frame
# a2=src, a3=dst, a4=len
mov a5, a3
@@ -303,6 +287,8 @@ bcopy:
mov a2, a5
j .Lmovecommon # go to common code for memmove+bcopy
+ENDPROC(bcopy)
+
/*
* void *memmove(void *dst, const void *src, size_t len);
*
@@ -391,10 +377,8 @@ bcopy:
j .Lbackdstaligned # dst is now aligned,
# return to main algorithm
- .align 4
- .global memmove
- .type memmove,@function
-memmove:
+ENTRY(__memmove)
+WEAK(memmove)
entry sp, 16 # minimal stack frame
# a2/ dst, a3/ src, a4/ len
@@ -485,7 +469,7 @@ memmove:
.Lbacksrcunaligned:
_beqz a4, .Lbackdone # avoid loading anything for zero-length copies
# copy 16 bytes per iteration for word-aligned dst and unaligned src
- ssa8 a3 # set shift amount from byte offset
+ __ssa8 a3 # set shift amount from byte offset
#define SIM_CHECKS_ALIGNMENT 1 /* set to 1 when running on ISS with
* the lint or ferret client, or 0
* to save a few cycles */
@@ -506,15 +490,15 @@ memmove:
l32i a7, a3, 12
l32i a8, a3, 8
addi a5, a5, -16
- src_b a6, a7, a6
+ __src_b a6, a7, a6
s32i a6, a5, 12
l32i a9, a3, 4
- src_b a7, a8, a7
+ __src_b a7, a8, a7
s32i a7, a5, 8
l32i a6, a3, 0
- src_b a8, a9, a8
+ __src_b a8, a9, a8
s32i a8, a5, 4
- src_b a9, a6, a9
+ __src_b a9, a6, a9
s32i a9, a5, 0
#if !XCHAL_HAVE_LOOPS
bne a3, a10, .backLoop2 # continue loop if a3:src != a10:src_start
@@ -526,9 +510,9 @@ memmove:
l32i a7, a3, 4
l32i a8, a3, 0
addi a5, a5, -8
- src_b a6, a7, a6
+ __src_b a6, a7, a6
s32i a6, a5, 4
- src_b a7, a8, a7
+ __src_b a7, a8, a7
s32i a7, a5, 0
mov a6, a8
.Lback12:
@@ -537,7 +521,7 @@ memmove:
addi a3, a3, -4
l32i a7, a3, 0
addi a5, a5, -4
- src_b a6, a7, a6
+ __src_b a6, a7, a6
s32i a6, a5, 0
mov a6, a7
.Lback13:
@@ -566,11 +550,4 @@ memmove:
s8i a6, a5, 0
retw
-
-/*
- * Local Variables:
- * mode:fundamental
- * comment-start: "# "
- * comment-start-skip: "# *"
- * End:
- */
+ENDPROC(__memmove)
diff --git a/arch/xtensa/lib/memset.S b/arch/xtensa/lib/memset.S
index 10b8c400f175..276747dec300 100644
--- a/arch/xtensa/lib/memset.S
+++ b/arch/xtensa/lib/memset.S
@@ -11,7 +11,9 @@
* Copyright (C) 2002 Tensilica Inc.
*/
+#include <linux/linkage.h>
#include <variant/core.h>
+#include <asm/asmmacro.h>
/*
* void *memset(void *dst, int c, size_t length)
@@ -28,20 +30,10 @@
* the alignment labels).
*/
-/* Load or store instructions that may cause exceptions use the EX macro. */
-
-#define EX(insn,reg1,reg2,offset,handler) \
-9: insn reg1, reg2, offset; \
- .section __ex_table, "a"; \
- .word 9b, handler; \
- .previous
-
-
.text
-.align 4
-.global memset
-.type memset,@function
-memset:
+ENTRY(__memset)
+WEAK(memset)
+
entry sp, 16 # minimal stack frame
# a2/ dst, a3/ c, a4/ length
extui a3, a3, 0, 8 # mask to just 8 bits
@@ -73,10 +65,10 @@ memset:
add a6, a6, a5 # a6 = end of last 16B chunk
#endif /* !XCHAL_HAVE_LOOPS */
.Loop1:
- EX(s32i, a3, a5, 0, memset_fixup)
- EX(s32i, a3, a5, 4, memset_fixup)
- EX(s32i, a3, a5, 8, memset_fixup)
- EX(s32i, a3, a5, 12, memset_fixup)
+EX(10f) s32i a3, a5, 0
+EX(10f) s32i a3, a5, 4
+EX(10f) s32i a3, a5, 8
+EX(10f) s32i a3, a5, 12
addi a5, a5, 16
#if !XCHAL_HAVE_LOOPS
blt a5, a6, .Loop1
@@ -84,23 +76,23 @@ memset:
.Loop1done:
bbci.l a4, 3, .L2
# set 8 bytes
- EX(s32i, a3, a5, 0, memset_fixup)
- EX(s32i, a3, a5, 4, memset_fixup)
+EX(10f) s32i a3, a5, 0
+EX(10f) s32i a3, a5, 4
addi a5, a5, 8
.L2:
bbci.l a4, 2, .L3
# set 4 bytes
- EX(s32i, a3, a5, 0, memset_fixup)
+EX(10f) s32i a3, a5, 0
addi a5, a5, 4
.L3:
bbci.l a4, 1, .L4
# set 2 bytes
- EX(s16i, a3, a5, 0, memset_fixup)
+EX(10f) s16i a3, a5, 0
addi a5, a5, 2
.L4:
bbci.l a4, 0, .L5
# set 1 byte
- EX(s8i, a3, a5, 0, memset_fixup)
+EX(10f) s8i a3, a5, 0
.L5:
.Lret1:
retw
@@ -114,7 +106,7 @@ memset:
bbci.l a5, 0, .L20 # branch if dst alignment half-aligned
# dst is only byte aligned
# set 1 byte
- EX(s8i, a3, a5, 0, memset_fixup)
+EX(10f) s8i a3, a5, 0
addi a5, a5, 1
addi a4, a4, -1
# now retest if dst aligned
@@ -122,7 +114,7 @@ memset:
.L20:
# dst half-aligned
# set 2 bytes
- EX(s16i, a3, a5, 0, memset_fixup)
+EX(10f) s16i a3, a5, 0
addi a5, a5, 2
addi a4, a4, -2
j .L0 # dst is now aligned, return to main algorithm
@@ -141,7 +133,7 @@ memset:
add a6, a5, a4 # a6 = ending address
#endif /* !XCHAL_HAVE_LOOPS */
.Lbyteloop:
- EX(s8i, a3, a5, 0, memset_fixup)
+EX(10f) s8i a3, a5, 0
addi a5, a5, 1
#if !XCHAL_HAVE_LOOPS
blt a5, a6, .Lbyteloop
@@ -149,12 +141,13 @@ memset:
.Lbytesetdone:
retw
+ENDPROC(__memset)
.section .fixup, "ax"
.align 4
/* We return zero if a failure occurred. */
-memset_fixup:
+10:
movi a2, 0
retw
diff --git a/arch/xtensa/lib/pci-auto.c b/arch/xtensa/lib/pci-auto.c
index 34d05abbd921..a2b558161d6d 100644
--- a/arch/xtensa/lib/pci-auto.c
+++ b/arch/xtensa/lib/pci-auto.c
@@ -49,17 +49,6 @@
*
*/
-
-/* define DEBUG to print some debugging messages. */
-
-#undef DEBUG
-
-#ifdef DEBUG
-# define DBG(x...) printk(x)
-#else
-# define DBG(x...)
-#endif
-
static int pciauto_upper_iospc;
static int pciauto_upper_memspc;
@@ -97,7 +86,7 @@ pciauto_setup_bars(struct pci_dev *dev, int bar_limit)
{
bar_size &= PCI_BASE_ADDRESS_IO_MASK;
upper_limit = &pciauto_upper_iospc;
- DBG("PCI Autoconfig: BAR %d, I/O, ", bar_nr);
+ pr_debug("PCI Autoconfig: BAR %d, I/O, ", bar_nr);
}
else
{
@@ -107,7 +96,7 @@ pciauto_setup_bars(struct pci_dev *dev, int bar_limit)
bar_size &= PCI_BASE_ADDRESS_MEM_MASK;
upper_limit = &pciauto_upper_memspc;
- DBG("PCI Autoconfig: BAR %d, Mem, ", bar_nr);
+ pr_debug("PCI Autoconfig: BAR %d, Mem, ", bar_nr);
}
/* Allocate a base address (bar_size is negative!) */
@@ -125,7 +114,8 @@ pciauto_setup_bars(struct pci_dev *dev, int bar_limit)
if (found_mem64)
pci_write_config_dword(dev, (bar+=4), 0x00000000);
- DBG("size=0x%x, address=0x%x\n", ~bar_size + 1, *upper_limit);
+ pr_debug("size=0x%x, address=0x%x\n",
+ ~bar_size + 1, *upper_limit);
}
}
@@ -150,7 +140,7 @@ pciauto_setup_irq(struct pci_controller* pci_ctrl,struct pci_dev *dev,int devfn)
if (irq == -1)
irq = 0;
- DBG("PCI Autoconfig: Interrupt %d, pin %d\n", irq, pin);
+ pr_debug("PCI Autoconfig: Interrupt %d, pin %d\n", irq, pin);
pci_write_config_byte(dev, PCI_INTERRUPT_LINE, irq);
}
@@ -289,8 +279,8 @@ int __init pciauto_bus_scan(struct pci_controller *pci_ctrl, int current_bus)
int iosave, memsave;
- DBG("PCI Autoconfig: Found P2P bridge, device %d\n",
- PCI_SLOT(pci_devfn));
+ pr_debug("PCI Autoconfig: Found P2P bridge, device %d\n",
+ PCI_SLOT(pci_devfn));
/* Allocate PCI I/O and/or memory space */
pciauto_setup_bars(dev, PCI_BASE_ADDRESS_1);
@@ -306,23 +296,6 @@ int __init pciauto_bus_scan(struct pci_controller *pci_ctrl, int current_bus)
}
-
-#if 0
- /* Skip legacy mode IDE controller */
-
- if ((pci_class >> 16) == PCI_CLASS_STORAGE_IDE) {
-
- unsigned char prg_iface;
- pci_read_config_byte(dev, PCI_CLASS_PROG, &prg_iface);
-
- if (!(prg_iface & PCIAUTO_IDE_MODE_MASK)) {
- DBG("PCI Autoconfig: Skipping legacy mode "
- "IDE controller\n");
- continue;
- }
- }
-#endif
-
/*
* Found a peripheral, enable some standard
* settings
@@ -337,8 +310,8 @@ int __init pciauto_bus_scan(struct pci_controller *pci_ctrl, int current_bus)
pci_write_config_byte(dev, PCI_LATENCY_TIMER, 0x80);
/* Allocate PCI I/O and/or memory space */
- DBG("PCI Autoconfig: Found Bus %d, Device %d, Function %d\n",
- current_bus, PCI_SLOT(pci_devfn), PCI_FUNC(pci_devfn) );
+ pr_debug("PCI Autoconfig: Found Bus %d, Device %d, Function %d\n",
+ current_bus, PCI_SLOT(pci_devfn), PCI_FUNC(pci_devfn));
pciauto_setup_bars(dev, PCI_BASE_ADDRESS_5);
pciauto_setup_irq(pci_ctrl, dev, pci_devfn);
diff --git a/arch/xtensa/lib/strncpy_user.S b/arch/xtensa/lib/strncpy_user.S
index 1ad0ecf45368..5fce16b67dca 100644
--- a/arch/xtensa/lib/strncpy_user.S
+++ b/arch/xtensa/lib/strncpy_user.S
@@ -11,16 +11,10 @@
* Copyright (C) 2002 Tensilica Inc.
*/
-#include <variant/core.h>
#include <linux/errno.h>
-
-/* Load or store instructions that may cause exceptions use the EX macro. */
-
-#define EX(insn,reg1,reg2,offset,handler) \
-9: insn reg1, reg2, offset; \
- .section __ex_table, "a"; \
- .word 9b, handler; \
- .previous
+#include <linux/linkage.h>
+#include <variant/core.h>
+#include <asm/asmmacro.h>
/*
* char *__strncpy_user(char *dst, const char *src, size_t len)
@@ -54,10 +48,8 @@
# a12/ tmp
.text
-.align 4
-.global __strncpy_user
-.type __strncpy_user,@function
-__strncpy_user:
+ENTRY(__strncpy_user)
+
entry sp, 16 # minimal stack frame
# a2/ dst, a3/ src, a4/ len
mov a11, a2 # leave dst in return value register
@@ -75,9 +67,9 @@ __strncpy_user:
j .Ldstunaligned
.Lsrc1mod2: # src address is odd
- EX(l8ui, a9, a3, 0, fixup_l) # get byte 0
+EX(11f) l8ui a9, a3, 0 # get byte 0
addi a3, a3, 1 # advance src pointer
- EX(s8i, a9, a11, 0, fixup_s) # store byte 0
+EX(10f) s8i a9, a11, 0 # store byte 0
beqz a9, .Lret # if byte 0 is zero
addi a11, a11, 1 # advance dst pointer
addi a4, a4, -1 # decrement len
@@ -85,16 +77,16 @@ __strncpy_user:
bbci.l a3, 1, .Lsrcaligned # if src is now word-aligned
.Lsrc2mod4: # src address is 2 mod 4
- EX(l8ui, a9, a3, 0, fixup_l) # get byte 0
+EX(11f) l8ui a9, a3, 0 # get byte 0
/* 1-cycle interlock */
- EX(s8i, a9, a11, 0, fixup_s) # store byte 0
+EX(10f) s8i a9, a11, 0 # store byte 0
beqz a9, .Lret # if byte 0 is zero
addi a11, a11, 1 # advance dst pointer
addi a4, a4, -1 # decrement len
beqz a4, .Lret # if len is zero
- EX(l8ui, a9, a3, 1, fixup_l) # get byte 0
+EX(11f) l8ui a9, a3, 1 # get byte 0
addi a3, a3, 2 # advance src pointer
- EX(s8i, a9, a11, 0, fixup_s) # store byte 0
+EX(10f) s8i a9, a11, 0 # store byte 0
beqz a9, .Lret # if byte 0 is zero
addi a11, a11, 1 # advance dst pointer
addi a4, a4, -1 # decrement len
@@ -117,12 +109,12 @@ __strncpy_user:
add a12, a12, a11 # a12 = end of last 4B chunck
#endif
.Loop1:
- EX(l32i, a9, a3, 0, fixup_l) # get word from src
+EX(11f) l32i a9, a3, 0 # get word from src
addi a3, a3, 4 # advance src pointer
bnone a9, a5, .Lz0 # if byte 0 is zero
bnone a9, a6, .Lz1 # if byte 1 is zero
bnone a9, a7, .Lz2 # if byte 2 is zero
- EX(s32i, a9, a11, 0, fixup_s) # store word to dst
+EX(10f) s32i a9, a11, 0 # store word to dst
bnone a9, a8, .Lz3 # if byte 3 is zero
addi a11, a11, 4 # advance dst pointer
#if !XCHAL_HAVE_LOOPS
@@ -132,7 +124,7 @@ __strncpy_user:
.Loop1done:
bbci.l a4, 1, .L100
# copy 2 bytes
- EX(l16ui, a9, a3, 0, fixup_l)
+EX(11f) l16ui a9, a3, 0
addi a3, a3, 2 # advance src pointer
#ifdef __XTENSA_EB__
bnone a9, a7, .Lz0 # if byte 2 is zero
@@ -141,13 +133,13 @@ __strncpy_user:
bnone a9, a5, .Lz0 # if byte 0 is zero
bnone a9, a6, .Lz1 # if byte 1 is zero
#endif
- EX(s16i, a9, a11, 0, fixup_s)
+EX(10f) s16i a9, a11, 0
addi a11, a11, 2 # advance dst pointer
.L100:
bbci.l a4, 0, .Lret
- EX(l8ui, a9, a3, 0, fixup_l)
+EX(11f) l8ui a9, a3, 0
/* slot */
- EX(s8i, a9, a11, 0, fixup_s)
+EX(10f) s8i a9, a11, 0
beqz a9, .Lret # if byte is zero
addi a11, a11, 1-3 # advance dst ptr 1, but also cancel
# the effect of adding 3 in .Lz3 code
@@ -161,14 +153,14 @@ __strncpy_user:
#ifdef __XTENSA_EB__
movi a9, 0
#endif /* __XTENSA_EB__ */
- EX(s8i, a9, a11, 0, fixup_s)
+EX(10f) s8i a9, a11, 0
sub a2, a11, a2 # compute strlen
retw
.Lz1: # byte 1 is zero
#ifdef __XTENSA_EB__
extui a9, a9, 16, 16
#endif /* __XTENSA_EB__ */
- EX(s16i, a9, a11, 0, fixup_s)
+EX(10f) s16i a9, a11, 0
addi a11, a11, 1 # advance dst pointer
sub a2, a11, a2 # compute strlen
retw
@@ -176,9 +168,9 @@ __strncpy_user:
#ifdef __XTENSA_EB__
extui a9, a9, 16, 16
#endif /* __XTENSA_EB__ */
- EX(s16i, a9, a11, 0, fixup_s)
+EX(10f) s16i a9, a11, 0
movi a9, 0
- EX(s8i, a9, a11, 2, fixup_s)
+EX(10f) s8i a9, a11, 2
addi a11, a11, 2 # advance dst pointer
sub a2, a11, a2 # compute strlen
retw
@@ -196,9 +188,9 @@ __strncpy_user:
add a12, a11, a4 # a12 = ending address
#endif /* XCHAL_HAVE_LOOPS */
.Lnextbyte:
- EX(l8ui, a9, a3, 0, fixup_l)
+EX(11f) l8ui a9, a3, 0
addi a3, a3, 1
- EX(s8i, a9, a11, 0, fixup_s)
+EX(10f) s8i a9, a11, 0
beqz a9, .Lunalignedend
addi a11, a11, 1
#if !XCHAL_HAVE_LOOPS
@@ -209,6 +201,7 @@ __strncpy_user:
sub a2, a11, a2 # compute strlen
retw
+ENDPROC(__strncpy_user)
.section .fixup, "ax"
.align 4
@@ -218,8 +211,7 @@ __strncpy_user:
* implementation in memset(). Thus, we differentiate between
* load/store fixups. */
-fixup_s:
-fixup_l:
+10:
+11:
movi a2, -EFAULT
retw
-
diff --git a/arch/xtensa/lib/strnlen_user.S b/arch/xtensa/lib/strnlen_user.S
index 4c03b1e581e9..0b956ce7f386 100644
--- a/arch/xtensa/lib/strnlen_user.S
+++ b/arch/xtensa/lib/strnlen_user.S
@@ -11,15 +11,9 @@
* Copyright (C) 2002 Tensilica Inc.
*/
+#include <linux/linkage.h>
#include <variant/core.h>
-
-/* Load or store instructions that may cause exceptions use the EX macro. */
-
-#define EX(insn,reg1,reg2,offset,handler) \
-9: insn reg1, reg2, offset; \
- .section __ex_table, "a"; \
- .word 9b, handler; \
- .previous
+#include <asm/asmmacro.h>
/*
* size_t __strnlen_user(const char *s, size_t len)
@@ -49,10 +43,8 @@
# a10/ tmp
.text
-.align 4
-.global __strnlen_user
-.type __strnlen_user,@function
-__strnlen_user:
+ENTRY(__strnlen_user)
+
entry sp, 16 # minimal stack frame
# a2/ s, a3/ len
addi a4, a2, -4 # because we overincrement at the end;
@@ -77,7 +69,7 @@ __strnlen_user:
add a10, a10, a4 # a10 = end of last 4B chunk
#endif /* XCHAL_HAVE_LOOPS */
.Loop:
- EX(l32i, a9, a4, 4, lenfixup) # get next word of string
+EX(10f) l32i a9, a4, 4 # get next word of string
addi a4, a4, 4 # advance string pointer
bnone a9, a5, .Lz0 # if byte 0 is zero
bnone a9, a6, .Lz1 # if byte 1 is zero
@@ -88,7 +80,7 @@ __strnlen_user:
#endif
.Ldone:
- EX(l32i, a9, a4, 4, lenfixup) # load 4 bytes for remaining checks
+EX(10f) l32i a9, a4, 4 # load 4 bytes for remaining checks
bbci.l a3, 1, .L100
# check two more bytes (bytes 0, 1 of word)
@@ -125,14 +117,14 @@ __strnlen_user:
retw
.L1mod2: # address is odd
- EX(l8ui, a9, a4, 4, lenfixup) # get byte 0
+EX(10f) l8ui a9, a4, 4 # get byte 0
addi a4, a4, 1 # advance string pointer
beqz a9, .Lz3 # if byte 0 is zero
bbci.l a4, 1, .Laligned # if string pointer is now word-aligned
.L2mod4: # address is 2 mod 4
addi a4, a4, 2 # advance ptr for aligned access
- EX(l32i, a9, a4, 0, lenfixup) # get word with first two bytes of string
+EX(10f) l32i a9, a4, 0 # get word with first two bytes of string
bnone a9, a7, .Lz2 # if byte 2 (of word, not string) is zero
bany a9, a8, .Laligned # if byte 3 (of word, not string) is nonzero
# byte 3 is zero
@@ -140,8 +132,10 @@ __strnlen_user:
sub a2, a4, a2 # subtract to get length
retw
+ENDPROC(__strnlen_user)
+
.section .fixup, "ax"
.align 4
-lenfixup:
+10:
movi a2, 0
retw
diff --git a/arch/xtensa/lib/usercopy.S b/arch/xtensa/lib/usercopy.S
index d9cd766bde3e..64ab1971324f 100644
--- a/arch/xtensa/lib/usercopy.S
+++ b/arch/xtensa/lib/usercopy.S
@@ -53,30 +53,13 @@
* a11/ original length
*/
+#include <linux/linkage.h>
#include <variant/core.h>
-
-#ifdef __XTENSA_EB__
-#define ALIGN(R, W0, W1) src R, W0, W1
-#define SSA8(R) ssa8b R
-#else
-#define ALIGN(R, W0, W1) src R, W1, W0
-#define SSA8(R) ssa8l R
-#endif
-
-/* Load or store instructions that may cause exceptions use the EX macro. */
-
-#define EX(insn,reg1,reg2,offset,handler) \
-9: insn reg1, reg2, offset; \
- .section __ex_table, "a"; \
- .word 9b, handler; \
- .previous
-
+#include <asm/asmmacro.h>
.text
- .align 4
- .global __xtensa_copy_user
- .type __xtensa_copy_user,@function
-__xtensa_copy_user:
+ENTRY(__xtensa_copy_user)
+
entry sp, 16 # minimal stack frame
# a2/ dst, a3/ src, a4/ len
mov a5, a2 # copy dst so that a2 is return value
@@ -89,7 +72,7 @@ __xtensa_copy_user:
# per iteration
movi a8, 3 # if source is also aligned,
bnone a3, a8, .Laligned # then use word copy
- SSA8( a3) # set shift amount from byte offset
+ __ssa8 a3 # set shift amount from byte offset
bnez a4, .Lsrcunaligned
movi a2, 0 # return success for len==0
retw
@@ -102,9 +85,9 @@ __xtensa_copy_user:
bltui a4, 7, .Lbytecopy # do short copies byte by byte
# copy 1 byte
- EX(l8ui, a6, a3, 0, fixup)
+EX(10f) l8ui a6, a3, 0
addi a3, a3, 1
- EX(s8i, a6, a5, 0, fixup)
+EX(10f) s8i a6, a5, 0
addi a5, a5, 1
addi a4, a4, -1
bbci.l a5, 1, .Ldstaligned # if dst is now aligned, then
@@ -112,11 +95,11 @@ __xtensa_copy_user:
.Ldst2mod4: # dst 16-bit aligned
# copy 2 bytes
bltui a4, 6, .Lbytecopy # do short copies byte by byte
- EX(l8ui, a6, a3, 0, fixup)
- EX(l8ui, a7, a3, 1, fixup)
+EX(10f) l8ui a6, a3, 0
+EX(10f) l8ui a7, a3, 1
addi a3, a3, 2
- EX(s8i, a6, a5, 0, fixup)
- EX(s8i, a7, a5, 1, fixup)
+EX(10f) s8i a6, a5, 0
+EX(10f) s8i a7, a5, 1
addi a5, a5, 2
addi a4, a4, -2
j .Ldstaligned # dst is now aligned, return to main algorithm
@@ -135,9 +118,9 @@ __xtensa_copy_user:
add a7, a3, a4 # a7 = end address for source
#endif /* !XCHAL_HAVE_LOOPS */
.Lnextbyte:
- EX(l8ui, a6, a3, 0, fixup)
+EX(10f) l8ui a6, a3, 0
addi a3, a3, 1
- EX(s8i, a6, a5, 0, fixup)
+EX(10f) s8i a6, a5, 0
addi a5, a5, 1
#if !XCHAL_HAVE_LOOPS
blt a3, a7, .Lnextbyte
@@ -161,15 +144,15 @@ __xtensa_copy_user:
add a8, a8, a3 # a8 = end of last 16B source chunk
#endif /* !XCHAL_HAVE_LOOPS */
.Loop1:
- EX(l32i, a6, a3, 0, fixup)
- EX(l32i, a7, a3, 4, fixup)
- EX(s32i, a6, a5, 0, fixup)
- EX(l32i, a6, a3, 8, fixup)
- EX(s32i, a7, a5, 4, fixup)
- EX(l32i, a7, a3, 12, fixup)
- EX(s32i, a6, a5, 8, fixup)
+EX(10f) l32i a6, a3, 0
+EX(10f) l32i a7, a3, 4
+EX(10f) s32i a6, a5, 0
+EX(10f) l32i a6, a3, 8
+EX(10f) s32i a7, a5, 4
+EX(10f) l32i a7, a3, 12
+EX(10f) s32i a6, a5, 8
addi a3, a3, 16
- EX(s32i, a7, a5, 12, fixup)
+EX(10f) s32i a7, a5, 12
addi a5, a5, 16
#if !XCHAL_HAVE_LOOPS
blt a3, a8, .Loop1
@@ -177,31 +160,31 @@ __xtensa_copy_user:
.Loop1done:
bbci.l a4, 3, .L2
# copy 8 bytes
- EX(l32i, a6, a3, 0, fixup)
- EX(l32i, a7, a3, 4, fixup)
+EX(10f) l32i a6, a3, 0
+EX(10f) l32i a7, a3, 4
addi a3, a3, 8
- EX(s32i, a6, a5, 0, fixup)
- EX(s32i, a7, a5, 4, fixup)
+EX(10f) s32i a6, a5, 0
+EX(10f) s32i a7, a5, 4
addi a5, a5, 8
.L2:
bbci.l a4, 2, .L3
# copy 4 bytes
- EX(l32i, a6, a3, 0, fixup)
+EX(10f) l32i a6, a3, 0
addi a3, a3, 4
- EX(s32i, a6, a5, 0, fixup)
+EX(10f) s32i a6, a5, 0
addi a5, a5, 4
.L3:
bbci.l a4, 1, .L4
# copy 2 bytes
- EX(l16ui, a6, a3, 0, fixup)
+EX(10f) l16ui a6, a3, 0
addi a3, a3, 2
- EX(s16i, a6, a5, 0, fixup)
+EX(10f) s16i a6, a5, 0
addi a5, a5, 2
.L4:
bbci.l a4, 0, .L5
# copy 1 byte
- EX(l8ui, a6, a3, 0, fixup)
- EX(s8i, a6, a5, 0, fixup)
+EX(10f) l8ui a6, a3, 0
+EX(10f) s8i a6, a5, 0
.L5:
movi a2, 0 # return success for len bytes copied
retw
@@ -217,7 +200,7 @@ __xtensa_copy_user:
# copy 16 bytes per iteration for word-aligned dst and unaligned src
and a10, a3, a8 # save unalignment offset for below
sub a3, a3, a10 # align a3 (to avoid sim warnings only; not needed for hardware)
- EX(l32i, a6, a3, 0, fixup) # load first word
+EX(10f) l32i a6, a3, 0 # load first word
#if XCHAL_HAVE_LOOPS
loopnez a7, .Loop2done
#else /* !XCHAL_HAVE_LOOPS */
@@ -226,19 +209,19 @@ __xtensa_copy_user:
add a12, a12, a3 # a12 = end of last 16B source chunk
#endif /* !XCHAL_HAVE_LOOPS */
.Loop2:
- EX(l32i, a7, a3, 4, fixup)
- EX(l32i, a8, a3, 8, fixup)
- ALIGN( a6, a6, a7)
- EX(s32i, a6, a5, 0, fixup)
- EX(l32i, a9, a3, 12, fixup)
- ALIGN( a7, a7, a8)
- EX(s32i, a7, a5, 4, fixup)
- EX(l32i, a6, a3, 16, fixup)
- ALIGN( a8, a8, a9)
- EX(s32i, a8, a5, 8, fixup)
+EX(10f) l32i a7, a3, 4
+EX(10f) l32i a8, a3, 8
+ __src_b a6, a6, a7
+EX(10f) s32i a6, a5, 0
+EX(10f) l32i a9, a3, 12
+ __src_b a7, a7, a8
+EX(10f) s32i a7, a5, 4
+EX(10f) l32i a6, a3, 16
+ __src_b a8, a8, a9
+EX(10f) s32i a8, a5, 8
addi a3, a3, 16
- ALIGN( a9, a9, a6)
- EX(s32i, a9, a5, 12, fixup)
+ __src_b a9, a9, a6
+EX(10f) s32i a9, a5, 12
addi a5, a5, 16
#if !XCHAL_HAVE_LOOPS
blt a3, a12, .Loop2
@@ -246,43 +229,44 @@ __xtensa_copy_user:
.Loop2done:
bbci.l a4, 3, .L12
# copy 8 bytes
- EX(l32i, a7, a3, 4, fixup)
- EX(l32i, a8, a3, 8, fixup)
- ALIGN( a6, a6, a7)
- EX(s32i, a6, a5, 0, fixup)
+EX(10f) l32i a7, a3, 4
+EX(10f) l32i a8, a3, 8
+ __src_b a6, a6, a7
+EX(10f) s32i a6, a5, 0
addi a3, a3, 8
- ALIGN( a7, a7, a8)
- EX(s32i, a7, a5, 4, fixup)
+ __src_b a7, a7, a8
+EX(10f) s32i a7, a5, 4
addi a5, a5, 8
mov a6, a8
.L12:
bbci.l a4, 2, .L13
# copy 4 bytes
- EX(l32i, a7, a3, 4, fixup)
+EX(10f) l32i a7, a3, 4
addi a3, a3, 4
- ALIGN( a6, a6, a7)
- EX(s32i, a6, a5, 0, fixup)
+ __src_b a6, a6, a7
+EX(10f) s32i a6, a5, 0
addi a5, a5, 4
mov a6, a7
.L13:
add a3, a3, a10 # readjust a3 with correct misalignment
bbci.l a4, 1, .L14
# copy 2 bytes
- EX(l8ui, a6, a3, 0, fixup)
- EX(l8ui, a7, a3, 1, fixup)
+EX(10f) l8ui a6, a3, 0
+EX(10f) l8ui a7, a3, 1
addi a3, a3, 2
- EX(s8i, a6, a5, 0, fixup)
- EX(s8i, a7, a5, 1, fixup)
+EX(10f) s8i a6, a5, 0
+EX(10f) s8i a7, a5, 1
addi a5, a5, 2
.L14:
bbci.l a4, 0, .L15
# copy 1 byte
- EX(l8ui, a6, a3, 0, fixup)
- EX(s8i, a6, a5, 0, fixup)
+EX(10f) l8ui a6, a3, 0
+EX(10f) s8i a6, a5, 0
.L15:
movi a2, 0 # return success for len bytes copied
retw
+ENDPROC(__xtensa_copy_user)
.section .fixup, "ax"
.align 4
@@ -294,7 +278,7 @@ __xtensa_copy_user:
*/
-fixup:
+10:
sub a2, a5, a2 /* a2 <-- bytes copied */
sub a2, a11, a2 /* a2 <-- bytes not copied */
retw
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-17 03:29:36 +0100