aboutsummaryrefslogtreecommitdiffstats
path: root/arch/ppc/lib
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 18:20:36 -0400
committerLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 18:20:36 -0400
commit1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch)
tree0bba044c4ce775e45a88a51686b5d9f90697ea9d /arch/ppc/lib
Linux-2.6.12-rc2v2.6.12-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!
Diffstat (limited to 'arch/ppc/lib')
-rw-r--r--arch/ppc/lib/Makefile9
-rw-r--r--arch/ppc/lib/checksum.S225
-rw-r--r--arch/ppc/lib/dec_and_lock.c46
-rw-r--r--arch/ppc/lib/div64.S58
-rw-r--r--arch/ppc/lib/locks.c190
-rw-r--r--arch/ppc/lib/rheap.c693
-rw-r--r--arch/ppc/lib/strcase.c23
-rw-r--r--arch/ppc/lib/string.S716
8 files changed, 1960 insertions, 0 deletions
diff --git a/arch/ppc/lib/Makefile b/arch/ppc/lib/Makefile
new file mode 100644
index 000000000000..1c380e67d435
--- /dev/null
+++ b/arch/ppc/lib/Makefile
@@ -0,0 +1,9 @@
1#
2# Makefile for ppc-specific library files..
3#
4
5obj-y := checksum.o string.o strcase.o dec_and_lock.o div64.o
6
7obj-$(CONFIG_SMP) += locks.o
8obj-$(CONFIG_8xx) += rheap.o
9obj-$(CONFIG_CPM2) += rheap.o
diff --git a/arch/ppc/lib/checksum.S b/arch/ppc/lib/checksum.S
new file mode 100644
index 000000000000..7874e8a80455
--- /dev/null
+++ b/arch/ppc/lib/checksum.S
@@ -0,0 +1,225 @@
1/*
2 * This file contains assembly-language implementations
3 * of IP-style 1's complement checksum routines.
4 *
5 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 *
12 * Severely hacked about by Paul Mackerras (paulus@cs.anu.edu.au).
13 */
14
15#include <linux/sys.h>
16#include <asm/processor.h>
17#include <asm/errno.h>
18#include <asm/ppc_asm.h>
19
20 .text
21
22/*
23 * ip_fast_csum(buf, len) -- Optimized for IP header
24 * len is in words and is always >= 5.
25 */
26_GLOBAL(ip_fast_csum)
27 lwz r0,0(r3)
28 lwzu r5,4(r3)
29 addic. r4,r4,-2
30 addc r0,r0,r5
31 mtctr r4
32 blelr-
331: lwzu r4,4(r3)
34 adde r0,r0,r4
35 bdnz 1b
36 addze r0,r0 /* add in final carry */
37 rlwinm r3,r0,16,0,31 /* fold two halves together */
38 add r3,r0,r3
39 not r3,r3
40 srwi r3,r3,16
41 blr
42
43/*
44 * Compute checksum of TCP or UDP pseudo-header:
45 * csum_tcpudp_magic(saddr, daddr, len, proto, sum)
46 */
47_GLOBAL(csum_tcpudp_magic)
48 rlwimi r5,r6,16,0,15 /* put proto in upper half of len */
49 addc r0,r3,r4 /* add 4 32-bit words together */
50 adde r0,r0,r5
51 adde r0,r0,r7
52 addze r0,r0 /* add in final carry */
53 rlwinm r3,r0,16,0,31 /* fold two halves together */
54 add r3,r0,r3
55 not r3,r3
56 srwi r3,r3,16
57 blr
58
59/*
60 * computes the checksum of a memory block at buff, length len,
61 * and adds in "sum" (32-bit)
62 *
63 * csum_partial(buff, len, sum)
64 */
65_GLOBAL(csum_partial)
66 addic r0,r5,0
67 subi r3,r3,4
68 srwi. r6,r4,2
69 beq 3f /* if we're doing < 4 bytes */
70 andi. r5,r3,2 /* Align buffer to longword boundary */
71 beq+ 1f
72 lhz r5,4(r3) /* do 2 bytes to get aligned */
73 addi r3,r3,2
74 subi r4,r4,2
75 addc r0,r0,r5
76 srwi. r6,r4,2 /* # words to do */
77 beq 3f
781: mtctr r6
792: lwzu r5,4(r3) /* the bdnz has zero overhead, so it should */
80 adde r0,r0,r5 /* be unnecessary to unroll this loop */
81 bdnz 2b
82 andi. r4,r4,3
833: cmpwi 0,r4,2
84 blt+ 4f
85 lhz r5,4(r3)
86 addi r3,r3,2
87 subi r4,r4,2
88 adde r0,r0,r5
894: cmpwi 0,r4,1
90 bne+ 5f
91 lbz r5,4(r3)
92 slwi r5,r5,8 /* Upper byte of word */
93 adde r0,r0,r5
945: addze r3,r0 /* add in final carry */
95 blr
96
97/*
98 * Computes the checksum of a memory block at src, length len,
99 * and adds in "sum" (32-bit), while copying the block to dst.
100 * If an access exception occurs on src or dst, it stores -EFAULT
101 * to *src_err or *dst_err respectively, and (for an error on
102 * src) zeroes the rest of dst.
103 *
104 * csum_partial_copy_generic(src, dst, len, sum, src_err, dst_err)
105 */
106_GLOBAL(csum_partial_copy_generic)
107 addic r0,r6,0
108 subi r3,r3,4
109 subi r4,r4,4
110 srwi. r6,r5,2
111 beq 3f /* if we're doing < 4 bytes */
112 andi. r9,r4,2 /* Align dst to longword boundary */
113 beq+ 1f
11481: lhz r6,4(r3) /* do 2 bytes to get aligned */
115 addi r3,r3,2
116 subi r5,r5,2
11791: sth r6,4(r4)
118 addi r4,r4,2
119 addc r0,r0,r6
120 srwi. r6,r5,2 /* # words to do */
121 beq 3f
1221: srwi. r6,r5,4 /* # groups of 4 words to do */
123 beq 10f
124 mtctr r6
12571: lwz r6,4(r3)
12672: lwz r9,8(r3)
12773: lwz r10,12(r3)
12874: lwzu r11,16(r3)
129 adde r0,r0,r6
13075: stw r6,4(r4)
131 adde r0,r0,r9
13276: stw r9,8(r4)
133 adde r0,r0,r10
13477: stw r10,12(r4)
135 adde r0,r0,r11
13678: stwu r11,16(r4)
137 bdnz 71b
13810: rlwinm. r6,r5,30,30,31 /* # words left to do */
139 beq 13f
140 mtctr r6
14182: lwzu r9,4(r3)
14292: stwu r9,4(r4)
143 adde r0,r0,r9
144 bdnz 82b
14513: andi. r5,r5,3
1463: cmpwi 0,r5,2
147 blt+ 4f
14883: lhz r6,4(r3)
149 addi r3,r3,2
150 subi r5,r5,2
15193: sth r6,4(r4)
152 addi r4,r4,2
153 adde r0,r0,r6
1544: cmpwi 0,r5,1
155 bne+ 5f
15684: lbz r6,4(r3)
15794: stb r6,4(r4)
158 slwi r6,r6,8 /* Upper byte of word */
159 adde r0,r0,r6
1605: addze r3,r0 /* add in final carry */
161 blr
162
163/* These shouldn't go in the fixup section, since that would
164 cause the ex_table addresses to get out of order. */
165
166src_error_4:
167 mfctr r6 /* update # bytes remaining from ctr */
168 rlwimi r5,r6,4,0,27
169 b 79f
170src_error_1:
171 li r6,0
172 subi r5,r5,2
17395: sth r6,4(r4)
174 addi r4,r4,2
17579: srwi. r6,r5,2
176 beq 3f
177 mtctr r6
178src_error_2:
179 li r6,0
18096: stwu r6,4(r4)
181 bdnz 96b
1823: andi. r5,r5,3
183 beq src_error
184src_error_3:
185 li r6,0
186 mtctr r5
187 addi r4,r4,3
18897: stbu r6,1(r4)
189 bdnz 97b
190src_error:
191 cmpwi 0,r7,0
192 beq 1f
193 li r6,-EFAULT
194 stw r6,0(r7)
1951: addze r3,r0
196 blr
197
198dst_error:
199 cmpwi 0,r8,0
200 beq 1f
201 li r6,-EFAULT
202 stw r6,0(r8)
2031: addze r3,r0
204 blr
205
206.section __ex_table,"a"
207 .long 81b,src_error_1
208 .long 91b,dst_error
209 .long 71b,src_error_4
210 .long 72b,src_error_4
211 .long 73b,src_error_4
212 .long 74b,src_error_4
213 .long 75b,dst_error
214 .long 76b,dst_error
215 .long 77b,dst_error
216 .long 78b,dst_error
217 .long 82b,src_error_2
218 .long 92b,dst_error
219 .long 83b,src_error_3
220 .long 93b,dst_error
221 .long 84b,src_error_3
222 .long 94b,dst_error
223 .long 95b,dst_error
224 .long 96b,dst_error
225 .long 97b,dst_error
diff --git a/arch/ppc/lib/dec_and_lock.c b/arch/ppc/lib/dec_and_lock.c
new file mode 100644
index 000000000000..4ee888070d91
--- /dev/null
+++ b/arch/ppc/lib/dec_and_lock.c
@@ -0,0 +1,46 @@
1#include <linux/module.h>
2#include <linux/spinlock.h>
3#include <asm/atomic.h>
4#include <asm/system.h>
5
6/*
7 * This is an implementation of the notion of "decrement a
8 * reference count, and return locked if it decremented to zero".
9 *
10 * This implementation can be used on any architecture that
11 * has a cmpxchg, and where atomic->value is an int holding
12 * the value of the atomic (i.e. the high bits aren't used
13 * for a lock or anything like that).
14 *
15 * N.B. ATOMIC_DEC_AND_LOCK gets defined in include/linux/spinlock.h
16 * if spinlocks are empty and thus atomic_dec_and_lock is defined
17 * to be atomic_dec_and_test - in that case we don't need it
18 * defined here as well.
19 */
20
21#ifndef ATOMIC_DEC_AND_LOCK
22int _atomic_dec_and_lock(atomic_t *atomic, spinlock_t *lock)
23{
24 int counter;
25 int newcount;
26
27 for (;;) {
28 counter = atomic_read(atomic);
29 newcount = counter - 1;
30 if (!newcount)
31 break; /* do it the slow way */
32
33 newcount = cmpxchg(&atomic->counter, counter, newcount);
34 if (newcount == counter)
35 return 0;
36 }
37
38 spin_lock(lock);
39 if (atomic_dec_and_test(atomic))
40 return 1;
41 spin_unlock(lock);
42 return 0;
43}
44
45EXPORT_SYMBOL(_atomic_dec_and_lock);
46#endif /* ATOMIC_DEC_AND_LOCK */
diff --git a/arch/ppc/lib/div64.S b/arch/ppc/lib/div64.S
new file mode 100644
index 000000000000..3527569e9926
--- /dev/null
+++ b/arch/ppc/lib/div64.S
@@ -0,0 +1,58 @@
1/*
2 * Divide a 64-bit unsigned number by a 32-bit unsigned number.
3 * This routine assumes that the top 32 bits of the dividend are
4 * non-zero to start with.
5 * On entry, r3 points to the dividend, which get overwritten with
6 * the 64-bit quotient, and r4 contains the divisor.
7 * On exit, r3 contains the remainder.
8 *
9 * Copyright (C) 2002 Paul Mackerras, IBM Corp.
10 *
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public License
13 * as published by the Free Software Foundation; either version
14 * 2 of the License, or (at your option) any later version.
15 */
16#include <asm/ppc_asm.h>
17#include <asm/processor.h>
18
19_GLOBAL(__div64_32)
20 lwz r5,0(r3) # get the dividend into r5/r6
21 lwz r6,4(r3)
22 cmplw r5,r4
23 li r7,0
24 li r8,0
25 blt 1f
26 divwu r7,r5,r4 # if dividend.hi >= divisor,
27 mullw r0,r7,r4 # quotient.hi = dividend.hi / divisor
28 subf. r5,r0,r5 # dividend.hi %= divisor
29 beq 3f
301: mr r11,r5 # here dividend.hi != 0
31 andis. r0,r5,0xc000
32 bne 2f
33 cntlzw r0,r5 # we are shifting the dividend right
34 li r10,-1 # to make it < 2^32, and shifting
35 srw r10,r10,r0 # the divisor right the same amount,
36 add r9,r4,r10 # rounding up (so the estimate cannot
37 andc r11,r6,r10 # ever be too large, only too small)
38 andc r9,r9,r10
39 or r11,r5,r11
40 rotlw r9,r9,r0
41 rotlw r11,r11,r0
42 divwu r11,r11,r9 # then we divide the shifted quantities
432: mullw r10,r11,r4 # to get an estimate of the quotient,
44 mulhwu r9,r11,r4 # multiply the estimate by the divisor,
45 subfc r6,r10,r6 # take the product from the divisor,
46 add r8,r8,r11 # and add the estimate to the accumulated
47 subfe. r5,r9,r5 # quotient
48 bne 1b
493: cmplw r6,r4
50 blt 4f
51 divwu r0,r6,r4 # perform the remaining 32-bit division
52 mullw r10,r0,r4 # and get the remainder
53 add r8,r8,r0
54 subf r6,r10,r6
554: stw r7,0(r3) # return the quotient in *r3
56 stw r8,4(r3)
57 mr r3,r6 # return the remainder in r3
58 blr
diff --git a/arch/ppc/lib/locks.c b/arch/ppc/lib/locks.c
new file mode 100644
index 000000000000..694163d696d8
--- /dev/null
+++ b/arch/ppc/lib/locks.c
@@ -0,0 +1,190 @@
1/*
2 * Locks for smp ppc
3 *
4 * Written by Cort Dougan (cort@cs.nmt.edu)
5 */
6
7#include <linux/config.h>
8#include <linux/sched.h>
9#include <linux/spinlock.h>
10#include <linux/module.h>
11#include <asm/ppc_asm.h>
12#include <asm/smp.h>
13
14#ifdef CONFIG_DEBUG_SPINLOCK
15
16#undef INIT_STUCK
17#define INIT_STUCK 200000000 /*0xffffffff*/
18
19/*
20 * Try to acquire a spinlock.
21 * Only does the stwcx. if the load returned 0 - the Programming
22 * Environments Manual suggests not doing unnecessary stcwx.'s
23 * since they may inhibit forward progress by other CPUs in getting
24 * a lock.
25 */
26static inline unsigned long __spin_trylock(volatile unsigned long *lock)
27{
28 unsigned long ret;
29
30 __asm__ __volatile__ ("\n\
311: lwarx %0,0,%1\n\
32 cmpwi 0,%0,0\n\
33 bne 2f\n"
34 PPC405_ERR77(0,%1)
35" stwcx. %2,0,%1\n\
36 bne- 1b\n\
37 isync\n\
382:"
39 : "=&r"(ret)
40 : "r"(lock), "r"(1)
41 : "cr0", "memory");
42
43 return ret;
44}
45
46void _raw_spin_lock(spinlock_t *lock)
47{
48 int cpu = smp_processor_id();
49 unsigned int stuck = INIT_STUCK;
50 while (__spin_trylock(&lock->lock)) {
51 while ((unsigned volatile long)lock->lock != 0) {
52 if (!--stuck) {
53 printk("_spin_lock(%p) CPU#%d NIP %p"
54 " holder: cpu %ld pc %08lX\n",
55 lock, cpu, __builtin_return_address(0),
56 lock->owner_cpu,lock->owner_pc);
57 stuck = INIT_STUCK;
58 /* steal the lock */
59 /*xchg_u32((void *)&lock->lock,0);*/
60 }
61 }
62 }
63 lock->owner_pc = (unsigned long)__builtin_return_address(0);
64 lock->owner_cpu = cpu;
65}
66EXPORT_SYMBOL(_raw_spin_lock);
67
68int _raw_spin_trylock(spinlock_t *lock)
69{
70 if (__spin_trylock(&lock->lock))
71 return 0;
72 lock->owner_cpu = smp_processor_id();
73 lock->owner_pc = (unsigned long)__builtin_return_address(0);
74 return 1;
75}
76EXPORT_SYMBOL(_raw_spin_trylock);
77
78void _raw_spin_unlock(spinlock_t *lp)
79{
80 if ( !lp->lock )
81 printk("_spin_unlock(%p): no lock cpu %d curr PC %p %s/%d\n",
82 lp, smp_processor_id(), __builtin_return_address(0),
83 current->comm, current->pid);
84 if ( lp->owner_cpu != smp_processor_id() )
85 printk("_spin_unlock(%p): cpu %d trying clear of cpu %d pc %lx val %lx\n",
86 lp, smp_processor_id(), (int)lp->owner_cpu,
87 lp->owner_pc,lp->lock);
88 lp->owner_pc = lp->owner_cpu = 0;
89 wmb();
90 lp->lock = 0;
91}
92EXPORT_SYMBOL(_raw_spin_unlock);
93
94/*
95 * For rwlocks, zero is unlocked, -1 is write-locked,
96 * positive is read-locked.
97 */
98static __inline__ int __read_trylock(rwlock_t *rw)
99{
100 signed int tmp;
101
102 __asm__ __volatile__(
103"2: lwarx %0,0,%1 # __read_trylock\n\
104 addic. %0,%0,1\n\
105 ble- 1f\n"
106 PPC405_ERR77(0,%1)
107" stwcx. %0,0,%1\n\
108 bne- 2b\n\
109 isync\n\
1101:"
111 : "=&r"(tmp)
112 : "r"(&rw->lock)
113 : "cr0", "memory");
114
115 return tmp;
116}
117
118int _raw_read_trylock(rwlock_t *rw)
119{
120 return __read_trylock(rw) > 0;
121}
122EXPORT_SYMBOL(_raw_read_trylock);
123
124void _raw_read_lock(rwlock_t *rw)
125{
126 unsigned int stuck;
127
128 while (__read_trylock(rw) <= 0) {
129 stuck = INIT_STUCK;
130 while (!read_can_lock(rw)) {
131 if (--stuck == 0) {
132 printk("_read_lock(%p) CPU#%d lock %d\n",
133 rw, _smp_processor_id(), rw->lock);
134 stuck = INIT_STUCK;
135 }
136 }
137 }
138}
139EXPORT_SYMBOL(_raw_read_lock);
140
141void _raw_read_unlock(rwlock_t *rw)
142{
143 if ( rw->lock == 0 )
144 printk("_read_unlock(): %s/%d (nip %08lX) lock %d\n",
145 current->comm,current->pid,current->thread.regs->nip,
146 rw->lock);
147 wmb();
148 atomic_dec((atomic_t *) &(rw)->lock);
149}
150EXPORT_SYMBOL(_raw_read_unlock);
151
152void _raw_write_lock(rwlock_t *rw)
153{
154 unsigned int stuck;
155
156 while (cmpxchg(&rw->lock, 0, -1) != 0) {
157 stuck = INIT_STUCK;
158 while (!write_can_lock(rw)) {
159 if (--stuck == 0) {
160 printk("write_lock(%p) CPU#%d lock %d)\n",
161 rw, _smp_processor_id(), rw->lock);
162 stuck = INIT_STUCK;
163 }
164 }
165 }
166 wmb();
167}
168EXPORT_SYMBOL(_raw_write_lock);
169
170int _raw_write_trylock(rwlock_t *rw)
171{
172 if (cmpxchg(&rw->lock, 0, -1) != 0)
173 return 0;
174 wmb();
175 return 1;
176}
177EXPORT_SYMBOL(_raw_write_trylock);
178
179void _raw_write_unlock(rwlock_t *rw)
180{
181 if (rw->lock >= 0)
182 printk("_write_lock(): %s/%d (nip %08lX) lock %d\n",
183 current->comm,current->pid,current->thread.regs->nip,
184 rw->lock);
185 wmb();
186 rw->lock = 0;
187}
188EXPORT_SYMBOL(_raw_write_unlock);
189
190#endif
diff --git a/arch/ppc/lib/rheap.c b/arch/ppc/lib/rheap.c
new file mode 100644
index 000000000000..42c5de2c898f
--- /dev/null
+++ b/arch/ppc/lib/rheap.c
@@ -0,0 +1,693 @@
1/*
2 * arch/ppc/syslib/rheap.c
3 *
4 * A Remote Heap. Remote means that we don't touch the memory that the
5 * heap points to. Normal heap implementations use the memory they manage
6 * to place their list. We cannot do that because the memory we manage may
7 * have special properties, for example it is uncachable or of different
8 * endianess.
9 *
10 * Author: Pantelis Antoniou <panto@intracom.gr>
11 *
12 * 2004 (c) INTRACOM S.A. Greece. This file is licensed under
13 * the terms of the GNU General Public License version 2. This program
14 * is licensed "as is" without any warranty of any kind, whether express
15 * or implied.
16 */
17#include <linux/types.h>
18#include <linux/errno.h>
19#include <linux/mm.h>
20#include <linux/slab.h>
21
22#include <asm/rheap.h>
23
24/*
25 * Fixup a list_head, needed when copying lists. If the pointers fall
26 * between s and e, apply the delta. This assumes that
27 * sizeof(struct list_head *) == sizeof(unsigned long *).
28 */
29static inline void fixup(unsigned long s, unsigned long e, int d,
30 struct list_head *l)
31{
32 unsigned long *pp;
33
34 pp = (unsigned long *)&l->next;
35 if (*pp >= s && *pp < e)
36 *pp += d;
37
38 pp = (unsigned long *)&l->prev;
39 if (*pp >= s && *pp < e)
40 *pp += d;
41}
42
43/* Grow the allocated blocks */
44static int grow(rh_info_t * info, int max_blocks)
45{
46 rh_block_t *block, *blk;
47 int i, new_blocks;
48 int delta;
49 unsigned long blks, blke;
50
51 if (max_blocks <= info->max_blocks)
52 return -EINVAL;
53
54 new_blocks = max_blocks - info->max_blocks;
55
56 block = kmalloc(sizeof(rh_block_t) * max_blocks, GFP_KERNEL);
57 if (block == NULL)
58 return -ENOMEM;
59
60 if (info->max_blocks > 0) {
61
62 /* copy old block area */
63 memcpy(block, info->block,
64 sizeof(rh_block_t) * info->max_blocks);
65
66 delta = (char *)block - (char *)info->block;
67
68 /* and fixup list pointers */
69 blks = (unsigned long)info->block;
70 blke = (unsigned long)(info->block + info->max_blocks);
71
72 for (i = 0, blk = block; i < info->max_blocks; i++, blk++)
73 fixup(blks, blke, delta, &blk->list);
74
75 fixup(blks, blke, delta, &info->empty_list);
76 fixup(blks, blke, delta, &info->free_list);
77 fixup(blks, blke, delta, &info->taken_list);
78
79 /* free the old allocated memory */
80 if ((info->flags & RHIF_STATIC_BLOCK) == 0)
81 kfree(info->block);
82 }
83
84 info->block = block;
85 info->empty_slots += new_blocks;
86 info->max_blocks = max_blocks;
87 info->flags &= ~RHIF_STATIC_BLOCK;
88
89 /* add all new blocks to the free list */
90 for (i = 0, blk = block + info->max_blocks; i < new_blocks; i++, blk++)
91 list_add(&blk->list, &info->empty_list);
92
93 return 0;
94}
95
96/*
97 * Assure at least the required amount of empty slots. If this function
98 * causes a grow in the block area then all pointers kept to the block
99 * area are invalid!
100 */
101static int assure_empty(rh_info_t * info, int slots)
102{
103 int max_blocks;
104
105 /* This function is not meant to be used to grow uncontrollably */
106 if (slots >= 4)
107 return -EINVAL;
108
109 /* Enough space */
110 if (info->empty_slots >= slots)
111 return 0;
112
113 /* Next 16 sized block */
114 max_blocks = ((info->max_blocks + slots) + 15) & ~15;
115
116 return grow(info, max_blocks);
117}
118
119static rh_block_t *get_slot(rh_info_t * info)
120{
121 rh_block_t *blk;
122
123 /* If no more free slots, and failure to extend. */
124 /* XXX: You should have called assure_empty before */
125 if (info->empty_slots == 0) {
126 printk(KERN_ERR "rh: out of slots; crash is imminent.\n");
127 return NULL;
128 }
129
130 /* Get empty slot to use */
131 blk = list_entry(info->empty_list.next, rh_block_t, list);
132 list_del_init(&blk->list);
133 info->empty_slots--;
134
135 /* Initialize */
136 blk->start = NULL;
137 blk->size = 0;
138 blk->owner = NULL;
139
140 return blk;
141}
142
143static inline void release_slot(rh_info_t * info, rh_block_t * blk)
144{
145 list_add(&blk->list, &info->empty_list);
146 info->empty_slots++;
147}
148
149static void attach_free_block(rh_info_t * info, rh_block_t * blkn)
150{
151 rh_block_t *blk;
152 rh_block_t *before;
153 rh_block_t *after;
154 rh_block_t *next;
155 int size;
156 unsigned long s, e, bs, be;
157 struct list_head *l;
158
159 /* We assume that they are aligned properly */
160 size = blkn->size;
161 s = (unsigned long)blkn->start;
162 e = s + size;
163
164 /* Find the blocks immediately before and after the given one
165 * (if any) */
166 before = NULL;
167 after = NULL;
168 next = NULL;
169
170 list_for_each(l, &info->free_list) {
171 blk = list_entry(l, rh_block_t, list);
172
173 bs = (unsigned long)blk->start;
174 be = bs + blk->size;
175
176 if (next == NULL && s >= bs)
177 next = blk;
178
179 if (be == s)
180 before = blk;
181
182 if (e == bs)
183 after = blk;
184
185 /* If both are not null, break now */
186 if (before != NULL && after != NULL)
187 break;
188 }
189
190 /* Now check if they are really adjacent */
191 if (before != NULL && s != (unsigned long)before->start + before->size)
192 before = NULL;
193
194 if (after != NULL && e != (unsigned long)after->start)
195 after = NULL;
196
197 /* No coalescing; list insert and return */
198 if (before == NULL && after == NULL) {
199
200 if (next != NULL)
201 list_add(&blkn->list, &next->list);
202 else
203 list_add(&blkn->list, &info->free_list);
204
205 return;
206 }
207
208 /* We don't need it anymore */
209 release_slot(info, blkn);
210
211 /* Grow the before block */
212 if (before != NULL && after == NULL) {
213 before->size += size;
214 return;
215 }
216
217 /* Grow the after block backwards */
218 if (before == NULL && after != NULL) {
219 after->start = (int8_t *)after->start - size;
220 after->size += size;
221 return;
222 }
223
224 /* Grow the before block, and release the after block */
225 before->size += size + after->size;
226 list_del(&after->list);
227 release_slot(info, after);
228}
229
230static void attach_taken_block(rh_info_t * info, rh_block_t * blkn)
231{
232 rh_block_t *blk;
233 struct list_head *l;
234
235 /* Find the block immediately before the given one (if any) */
236 list_for_each(l, &info->taken_list) {
237 blk = list_entry(l, rh_block_t, list);
238 if (blk->start > blkn->start) {
239 list_add_tail(&blkn->list, &blk->list);
240 return;
241 }
242 }
243
244 list_add_tail(&blkn->list, &info->taken_list);
245}
246
247/*
248 * Create a remote heap dynamically. Note that no memory for the blocks
249 * are allocated. It will upon the first allocation
250 */
251rh_info_t *rh_create(unsigned int alignment)
252{
253 rh_info_t *info;
254
255 /* Alignment must be a power of two */
256 if ((alignment & (alignment - 1)) != 0)
257 return ERR_PTR(-EINVAL);
258
259 info = kmalloc(sizeof(*info), GFP_KERNEL);
260 if (info == NULL)
261 return ERR_PTR(-ENOMEM);
262
263 info->alignment = alignment;
264
265 /* Initially everything as empty */
266 info->block = NULL;
267 info->max_blocks = 0;
268 info->empty_slots = 0;
269 info->flags = 0;
270
271 INIT_LIST_HEAD(&info->empty_list);
272 INIT_LIST_HEAD(&info->free_list);
273 INIT_LIST_HEAD(&info->taken_list);
274
275 return info;
276}
277
278/*
279 * Destroy a dynamically created remote heap. Deallocate only if the areas
280 * are not static
281 */
282void rh_destroy(rh_info_t * info)
283{
284 if ((info->flags & RHIF_STATIC_BLOCK) == 0 && info->block != NULL)
285 kfree(info->block);
286
287 if ((info->flags & RHIF_STATIC_INFO) == 0)
288 kfree(info);
289}
290
291/*
292 * Initialize in place a remote heap info block. This is needed to support
293 * operation very early in the startup of the kernel, when it is not yet safe
294 * to call kmalloc.
295 */
296void rh_init(rh_info_t * info, unsigned int alignment, int max_blocks,
297 rh_block_t * block)
298{
299 int i;
300 rh_block_t *blk;
301
302 /* Alignment must be a power of two */
303 if ((alignment & (alignment - 1)) != 0)
304 return;
305
306 info->alignment = alignment;
307
308 /* Initially everything as empty */
309 info->block = block;
310 info->max_blocks = max_blocks;
311 info->empty_slots = max_blocks;
312 info->flags = RHIF_STATIC_INFO | RHIF_STATIC_BLOCK;
313
314 INIT_LIST_HEAD(&info->empty_list);
315 INIT_LIST_HEAD(&info->free_list);
316 INIT_LIST_HEAD(&info->taken_list);
317
318 /* Add all new blocks to the free list */
319 for (i = 0, blk = block; i < max_blocks; i++, blk++)
320 list_add(&blk->list, &info->empty_list);
321}
322
323/* Attach a free memory region, coalesces regions if adjuscent */
324int rh_attach_region(rh_info_t * info, void *start, int size)
325{
326 rh_block_t *blk;
327 unsigned long s, e, m;
328 int r;
329
330 /* The region must be aligned */
331 s = (unsigned long)start;
332 e = s + size;
333 m = info->alignment - 1;
334
335 /* Round start up */
336 s = (s + m) & ~m;
337
338 /* Round end down */
339 e = e & ~m;
340
341 /* Take final values */
342 start = (void *)s;
343 size = (int)(e - s);
344
345 /* Grow the blocks, if needed */
346 r = assure_empty(info, 1);
347 if (r < 0)
348 return r;
349
350 blk = get_slot(info);
351 blk->start = start;
352 blk->size = size;
353 blk->owner = NULL;
354
355 attach_free_block(info, blk);
356
357 return 0;
358}
359
360/* Detatch given address range, splits free block if needed. */
361void *rh_detach_region(rh_info_t * info, void *start, int size)
362{
363 struct list_head *l;
364 rh_block_t *blk, *newblk;
365 unsigned long s, e, m, bs, be;
366
367 /* Validate size */
368 if (size <= 0)
369 return ERR_PTR(-EINVAL);
370
371 /* The region must be aligned */
372 s = (unsigned long)start;
373 e = s + size;
374 m = info->alignment - 1;
375
376 /* Round start up */
377 s = (s + m) & ~m;
378
379 /* Round end down */
380 e = e & ~m;
381
382 if (assure_empty(info, 1) < 0)
383 return ERR_PTR(-ENOMEM);
384
385 blk = NULL;
386 list_for_each(l, &info->free_list) {
387 blk = list_entry(l, rh_block_t, list);
388 /* The range must lie entirely inside one free block */
389 bs = (unsigned long)blk->start;
390 be = (unsigned long)blk->start + blk->size;
391 if (s >= bs && e <= be)
392 break;
393 blk = NULL;
394 }
395
396 if (blk == NULL)
397 return ERR_PTR(-ENOMEM);
398
399 /* Perfect fit */
400 if (bs == s && be == e) {
401 /* Delete from free list, release slot */
402 list_del(&blk->list);
403 release_slot(info, blk);
404 return (void *)s;
405 }
406
407 /* blk still in free list, with updated start and/or size */
408 if (bs == s || be == e) {
409 if (bs == s)
410 blk->start = (int8_t *)blk->start + size;
411 blk->size -= size;
412
413 } else {
414 /* The front free fragment */
415 blk->size = s - bs;
416
417 /* the back free fragment */
418 newblk = get_slot(info);
419 newblk->start = (void *)e;
420 newblk->size = be - e;
421
422 list_add(&newblk->list, &blk->list);
423 }
424
425 return (void *)s;
426}
427
428void *rh_alloc(rh_info_t * info, int size, const char *owner)
429{
430 struct list_head *l;
431 rh_block_t *blk;
432 rh_block_t *newblk;
433 void *start;
434
435 /* Validate size */
436 if (size <= 0)
437 return ERR_PTR(-EINVAL);
438
439 /* Align to configured alignment */
440 size = (size + (info->alignment - 1)) & ~(info->alignment - 1);
441
442 if (assure_empty(info, 1) < 0)
443 return ERR_PTR(-ENOMEM);
444
445 blk = NULL;
446 list_for_each(l, &info->free_list) {
447 blk = list_entry(l, rh_block_t, list);
448 if (size <= blk->size)
449 break;
450 blk = NULL;
451 }
452
453 if (blk == NULL)
454 return ERR_PTR(-ENOMEM);
455
456 /* Just fits */
457 if (blk->size == size) {
458 /* Move from free list to taken list */
459 list_del(&blk->list);
460 blk->owner = owner;
461 start = blk->start;
462
463 attach_taken_block(info, blk);
464
465 return start;
466 }
467
468 newblk = get_slot(info);
469 newblk->start = blk->start;
470 newblk->size = size;
471 newblk->owner = owner;
472
473 /* blk still in free list, with updated start, size */
474 blk->start = (int8_t *)blk->start + size;
475 blk->size -= size;
476
477 start = newblk->start;
478
479 attach_taken_block(info, newblk);
480
481 return start;
482}
483
484/* allocate at precisely the given address */
485void *rh_alloc_fixed(rh_info_t * info, void *start, int size, const char *owner)
486{
487 struct list_head *l;
488 rh_block_t *blk, *newblk1, *newblk2;
489 unsigned long s, e, m, bs, be;
490
491 /* Validate size */
492 if (size <= 0)
493 return ERR_PTR(-EINVAL);
494
495 /* The region must be aligned */
496 s = (unsigned long)start;
497 e = s + size;
498 m = info->alignment - 1;
499
500 /* Round start up */
501 s = (s + m) & ~m;
502
503 /* Round end down */
504 e = e & ~m;
505
506 if (assure_empty(info, 2) < 0)
507 return ERR_PTR(-ENOMEM);
508
509 blk = NULL;
510 list_for_each(l, &info->free_list) {
511 blk = list_entry(l, rh_block_t, list);
512 /* The range must lie entirely inside one free block */
513 bs = (unsigned long)blk->start;
514 be = (unsigned long)blk->start + blk->size;
515 if (s >= bs && e <= be)
516 break;
517 }
518
519 if (blk == NULL)
520 return ERR_PTR(-ENOMEM);
521
522 /* Perfect fit */
523 if (bs == s && be == e) {
524 /* Move from free list to taken list */
525 list_del(&blk->list);
526 blk->owner = owner;
527
528 start = blk->start;
529 attach_taken_block(info, blk);
530
531 return start;
532
533 }
534
535 /* blk still in free list, with updated start and/or size */
536 if (bs == s || be == e) {
537 if (bs == s)
538 blk->start = (int8_t *)blk->start + size;
539 blk->size -= size;
540
541 } else {
542 /* The front free fragment */
543 blk->size = s - bs;
544
545 /* The back free fragment */
546 newblk2 = get_slot(info);
547 newblk2->start = (void *)e;
548 newblk2->size = be - e;
549
550 list_add(&newblk2->list, &blk->list);
551 }
552
553 newblk1 = get_slot(info);
554 newblk1->start = (void *)s;
555 newblk1->size = e - s;
556 newblk1->owner = owner;
557
558 start = newblk1->start;
559 attach_taken_block(info, newblk1);
560
561 return start;
562}
563
564int rh_free(rh_info_t * info, void *start)
565{
566 rh_block_t *blk, *blk2;
567 struct list_head *l;
568 int size;
569
570 /* Linear search for block */
571 blk = NULL;
572 list_for_each(l, &info->taken_list) {
573 blk2 = list_entry(l, rh_block_t, list);
574 if (start < blk2->start)
575 break;
576 blk = blk2;
577 }
578
579 if (blk == NULL || start > (blk->start + blk->size))
580 return -EINVAL;
581
582 /* Remove from taken list */
583 list_del(&blk->list);
584
585 /* Get size of freed block */
586 size = blk->size;
587 attach_free_block(info, blk);
588
589 return size;
590}
591
592int rh_get_stats(rh_info_t * info, int what, int max_stats, rh_stats_t * stats)
593{
594 rh_block_t *blk;
595 struct list_head *l;
596 struct list_head *h;
597 int nr;
598
599 switch (what) {
600
601 case RHGS_FREE:
602 h = &info->free_list;
603 break;
604
605 case RHGS_TAKEN:
606 h = &info->taken_list;
607 break;
608
609 default:
610 return -EINVAL;
611 }
612
613 /* Linear search for block */
614 nr = 0;
615 list_for_each(l, h) {
616 blk = list_entry(l, rh_block_t, list);
617 if (stats != NULL && nr < max_stats) {
618 stats->start = blk->start;
619 stats->size = blk->size;
620 stats->owner = blk->owner;
621 stats++;
622 }
623 nr++;
624 }
625
626 return nr;
627}
628
629int rh_set_owner(rh_info_t * info, void *start, const char *owner)
630{
631 rh_block_t *blk, *blk2;
632 struct list_head *l;
633 int size;
634
635 /* Linear search for block */
636 blk = NULL;
637 list_for_each(l, &info->taken_list) {
638 blk2 = list_entry(l, rh_block_t, list);
639 if (start < blk2->start)
640 break;
641 blk = blk2;
642 }
643
644 if (blk == NULL || start > (blk->start + blk->size))
645 return -EINVAL;
646
647 blk->owner = owner;
648 size = blk->size;
649
650 return size;
651}
652
653void rh_dump(rh_info_t * info)
654{
655 static rh_stats_t st[32]; /* XXX maximum 32 blocks */
656 int maxnr;
657 int i, nr;
658
659 maxnr = sizeof(st) / sizeof(st[0]);
660
661 printk(KERN_INFO
662 "info @0x%p (%d slots empty / %d max)\n",
663 info, info->empty_slots, info->max_blocks);
664
665 printk(KERN_INFO " Free:\n");
666 nr = rh_get_stats(info, RHGS_FREE, maxnr, st);
667 if (nr > maxnr)
668 nr = maxnr;
669 for (i = 0; i < nr; i++)
670 printk(KERN_INFO
671 " 0x%p-0x%p (%u)\n",
672 st[i].start, (int8_t *) st[i].start + st[i].size,
673 st[i].size);
674 printk(KERN_INFO "\n");
675
676 printk(KERN_INFO " Taken:\n");
677 nr = rh_get_stats(info, RHGS_TAKEN, maxnr, st);
678 if (nr > maxnr)
679 nr = maxnr;
680 for (i = 0; i < nr; i++)
681 printk(KERN_INFO
682 " 0x%p-0x%p (%u) %s\n",
683 st[i].start, (int8_t *) st[i].start + st[i].size,
684 st[i].size, st[i].owner != NULL ? st[i].owner : "");
685 printk(KERN_INFO "\n");
686}
687
688void rh_dump_blk(rh_info_t * info, rh_block_t * blk)
689{
690 printk(KERN_INFO
691 "blk @0x%p: 0x%p-0x%p (%u)\n",
692 blk, blk->start, (int8_t *) blk->start + blk->size, blk->size);
693}
diff --git a/arch/ppc/lib/strcase.c b/arch/ppc/lib/strcase.c
new file mode 100644
index 000000000000..36b521091bbc
--- /dev/null
+++ b/arch/ppc/lib/strcase.c
@@ -0,0 +1,23 @@
1#include <linux/ctype.h>
2
3int strcasecmp(const char *s1, const char *s2)
4{
5 int c1, c2;
6
7 do {
8 c1 = tolower(*s1++);
9 c2 = tolower(*s2++);
10 } while (c1 == c2 && c1 != 0);
11 return c1 - c2;
12}
13
14int strncasecmp(const char *s1, const char *s2, int n)
15{
16 int c1, c2;
17
18 do {
19 c1 = tolower(*s1++);
20 c2 = tolower(*s2++);
21 } while ((--n > 0) && c1 == c2 && c1 != 0);
22 return c1 - c2;
23}
diff --git a/arch/ppc/lib/string.S b/arch/ppc/lib/string.S
new file mode 100644
index 000000000000..8d08a2eb225e
--- /dev/null
+++ b/arch/ppc/lib/string.S
@@ -0,0 +1,716 @@
1/*
2 * String handling functions for PowerPC.
3 *
4 * Copyright (C) 1996 Paul Mackerras.
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 */
11#include <linux/config.h>
12#include <asm/processor.h>
13#include <asm/cache.h>
14#include <asm/errno.h>
15#include <asm/ppc_asm.h>
16
17#define COPY_16_BYTES \
18 lwz r7,4(r4); \
19 lwz r8,8(r4); \
20 lwz r9,12(r4); \
21 lwzu r10,16(r4); \
22 stw r7,4(r6); \
23 stw r8,8(r6); \
24 stw r9,12(r6); \
25 stwu r10,16(r6)
26
27#define COPY_16_BYTES_WITHEX(n) \
288 ## n ## 0: \
29 lwz r7,4(r4); \
308 ## n ## 1: \
31 lwz r8,8(r4); \
328 ## n ## 2: \
33 lwz r9,12(r4); \
348 ## n ## 3: \
35 lwzu r10,16(r4); \
368 ## n ## 4: \
37 stw r7,4(r6); \
388 ## n ## 5: \
39 stw r8,8(r6); \
408 ## n ## 6: \
41 stw r9,12(r6); \
428 ## n ## 7: \
43 stwu r10,16(r6)
44
45#define COPY_16_BYTES_EXCODE(n) \
469 ## n ## 0: \
47 addi r5,r5,-(16 * n); \
48 b 104f; \
499 ## n ## 1: \
50 addi r5,r5,-(16 * n); \
51 b 105f; \
52.section __ex_table,"a"; \
53 .align 2; \
54 .long 8 ## n ## 0b,9 ## n ## 0b; \
55 .long 8 ## n ## 1b,9 ## n ## 0b; \
56 .long 8 ## n ## 2b,9 ## n ## 0b; \
57 .long 8 ## n ## 3b,9 ## n ## 0b; \
58 .long 8 ## n ## 4b,9 ## n ## 1b; \
59 .long 8 ## n ## 5b,9 ## n ## 1b; \
60 .long 8 ## n ## 6b,9 ## n ## 1b; \
61 .long 8 ## n ## 7b,9 ## n ## 1b; \
62 .text
63
64 .text
65 .stabs "arch/ppc/lib/",N_SO,0,0,0f
66 .stabs "string.S",N_SO,0,0,0f
67
68CACHELINE_BYTES = L1_CACHE_LINE_SIZE
69LG_CACHELINE_BYTES = LG_L1_CACHE_LINE_SIZE
70CACHELINE_MASK = (L1_CACHE_LINE_SIZE-1)
71
72_GLOBAL(strcpy)
73 addi r5,r3,-1
74 addi r4,r4,-1
751: lbzu r0,1(r4)
76 cmpwi 0,r0,0
77 stbu r0,1(r5)
78 bne 1b
79 blr
80
81/* This clears out any unused part of the destination buffer,
82 just as the libc version does. -- paulus */
83_GLOBAL(strncpy)
84 cmpwi 0,r5,0
85 beqlr
86 mtctr r5
87 addi r6,r3,-1
88 addi r4,r4,-1
891: lbzu r0,1(r4)
90 cmpwi 0,r0,0
91 stbu r0,1(r6)
92 bdnzf 2,1b /* dec ctr, branch if ctr != 0 && !cr0.eq */
93 bnelr /* if we didn't hit a null char, we're done */
94 mfctr r5
95 cmpwi 0,r5,0 /* any space left in destination buffer? */
96 beqlr /* we know r0 == 0 here */
972: stbu r0,1(r6) /* clear it out if so */
98 bdnz 2b
99 blr
100
101_GLOBAL(strcat)
102 addi r5,r3,-1
103 addi r4,r4,-1
1041: lbzu r0,1(r5)
105 cmpwi 0,r0,0
106 bne 1b
107 addi r5,r5,-1
1081: lbzu r0,1(r4)
109 cmpwi 0,r0,0
110 stbu r0,1(r5)
111 bne 1b
112 blr
113
114_GLOBAL(strcmp)
115 addi r5,r3,-1
116 addi r4,r4,-1
1171: lbzu r3,1(r5)
118 cmpwi 1,r3,0
119 lbzu r0,1(r4)
120 subf. r3,r0,r3
121 beqlr 1
122 beq 1b
123 blr
124
125_GLOBAL(strlen)
126 addi r4,r3,-1
1271: lbzu r0,1(r4)
128 cmpwi 0,r0,0
129 bne 1b
130 subf r3,r3,r4
131 blr
132
133/*
134 * Use dcbz on the complete cache lines in the destination
135 * to set them to zero. This requires that the destination
136 * area is cacheable. -- paulus
137 */
138_GLOBAL(cacheable_memzero)
139 mr r5,r4
140 li r4,0
141 addi r6,r3,-4
142 cmplwi 0,r5,4
143 blt 7f
144 stwu r4,4(r6)
145 beqlr
146 andi. r0,r6,3
147 add r5,r0,r5
148 subf r6,r0,r6
149 clrlwi r7,r6,32-LG_CACHELINE_BYTES
150 add r8,r7,r5
151 srwi r9,r8,LG_CACHELINE_BYTES
152 addic. r9,r9,-1 /* total number of complete cachelines */
153 ble 2f
154 xori r0,r7,CACHELINE_MASK & ~3
155 srwi. r0,r0,2
156 beq 3f
157 mtctr r0
1584: stwu r4,4(r6)
159 bdnz 4b
1603: mtctr r9
161 li r7,4
162#if !defined(CONFIG_8xx)
16310: dcbz r7,r6
164#else
16510: stw r4, 4(r6)
166 stw r4, 8(r6)
167 stw r4, 12(r6)
168 stw r4, 16(r6)
169#if CACHE_LINE_SIZE >= 32
170 stw r4, 20(r6)
171 stw r4, 24(r6)
172 stw r4, 28(r6)
173 stw r4, 32(r6)
174#endif /* CACHE_LINE_SIZE */
175#endif
176 addi r6,r6,CACHELINE_BYTES
177 bdnz 10b
178 clrlwi r5,r8,32-LG_CACHELINE_BYTES
179 addi r5,r5,4
1802: srwi r0,r5,2
181 mtctr r0
182 bdz 6f
1831: stwu r4,4(r6)
184 bdnz 1b
1856: andi. r5,r5,3
1867: cmpwi 0,r5,0
187 beqlr
188 mtctr r5
189 addi r6,r6,3
1908: stbu r4,1(r6)
191 bdnz 8b
192 blr
193
194_GLOBAL(memset)
195 rlwimi r4,r4,8,16,23
196 rlwimi r4,r4,16,0,15
197 addi r6,r3,-4
198 cmplwi 0,r5,4
199 blt 7f
200 stwu r4,4(r6)
201 beqlr
202 andi. r0,r6,3
203 add r5,r0,r5
204 subf r6,r0,r6
205 srwi r0,r5,2
206 mtctr r0
207 bdz 6f
2081: stwu r4,4(r6)
209 bdnz 1b
2106: andi. r5,r5,3
2117: cmpwi 0,r5,0
212 beqlr
213 mtctr r5
214 addi r6,r6,3
2158: stbu r4,1(r6)
216 bdnz 8b
217 blr
218
219/*
220 * This version uses dcbz on the complete cache lines in the
221 * destination area to reduce memory traffic. This requires that
222 * the destination area is cacheable.
223 * We only use this version if the source and dest don't overlap.
224 * -- paulus.
225 */
226_GLOBAL(cacheable_memcpy)
227 add r7,r3,r5 /* test if the src & dst overlap */
228 add r8,r4,r5
229 cmplw 0,r4,r7
230 cmplw 1,r3,r8
231 crand 0,0,4 /* cr0.lt &= cr1.lt */
232 blt memcpy /* if regions overlap */
233
234 addi r4,r4,-4
235 addi r6,r3,-4
236 neg r0,r3
237 andi. r0,r0,CACHELINE_MASK /* # bytes to start of cache line */
238 beq 58f
239
240 cmplw 0,r5,r0 /* is this more than total to do? */
241 blt 63f /* if not much to do */
242 andi. r8,r0,3 /* get it word-aligned first */
243 subf r5,r0,r5
244 mtctr r8
245 beq+ 61f
24670: lbz r9,4(r4) /* do some bytes */
247 stb r9,4(r6)
248 addi r4,r4,1
249 addi r6,r6,1
250 bdnz 70b
25161: srwi. r0,r0,2
252 mtctr r0
253 beq 58f
25472: lwzu r9,4(r4) /* do some words */
255 stwu r9,4(r6)
256 bdnz 72b
257
25858: srwi. r0,r5,LG_CACHELINE_BYTES /* # complete cachelines */
259 clrlwi r5,r5,32-LG_CACHELINE_BYTES
260 li r11,4
261 mtctr r0
262 beq 63f
26353:
264#if !defined(CONFIG_8xx)
265 dcbz r11,r6
266#endif
267 COPY_16_BYTES
268#if L1_CACHE_LINE_SIZE >= 32
269 COPY_16_BYTES
270#if L1_CACHE_LINE_SIZE >= 64
271 COPY_16_BYTES
272 COPY_16_BYTES
273#if L1_CACHE_LINE_SIZE >= 128
274 COPY_16_BYTES
275 COPY_16_BYTES
276 COPY_16_BYTES
277 COPY_16_BYTES
278#endif
279#endif
280#endif
281 bdnz 53b
282
28363: srwi. r0,r5,2
284 mtctr r0
285 beq 64f
28630: lwzu r0,4(r4)
287 stwu r0,4(r6)
288 bdnz 30b
289
29064: andi. r0,r5,3
291 mtctr r0
292 beq+ 65f
29340: lbz r0,4(r4)
294 stb r0,4(r6)
295 addi r4,r4,1
296 addi r6,r6,1
297 bdnz 40b
29865: blr
299
300_GLOBAL(memmove)
301 cmplw 0,r3,r4
302 bgt backwards_memcpy
303 /* fall through */
304
305_GLOBAL(memcpy)
306 srwi. r7,r5,3
307 addi r6,r3,-4
308 addi r4,r4,-4
309 beq 2f /* if less than 8 bytes to do */
310 andi. r0,r6,3 /* get dest word aligned */
311 mtctr r7
312 bne 5f
3131: lwz r7,4(r4)
314 lwzu r8,8(r4)
315 stw r7,4(r6)
316 stwu r8,8(r6)
317 bdnz 1b
318 andi. r5,r5,7
3192: cmplwi 0,r5,4
320 blt 3f
321 lwzu r0,4(r4)
322 addi r5,r5,-4
323 stwu r0,4(r6)
3243: cmpwi 0,r5,0
325 beqlr
326 mtctr r5
327 addi r4,r4,3
328 addi r6,r6,3
3294: lbzu r0,1(r4)
330 stbu r0,1(r6)
331 bdnz 4b
332 blr
3335: subfic r0,r0,4
334 mtctr r0
3356: lbz r7,4(r4)
336 addi r4,r4,1
337 stb r7,4(r6)
338 addi r6,r6,1
339 bdnz 6b
340 subf r5,r0,r5
341 rlwinm. r7,r5,32-3,3,31
342 beq 2b
343 mtctr r7
344 b 1b
345
346_GLOBAL(backwards_memcpy)
347 rlwinm. r7,r5,32-3,3,31 /* r0 = r5 >> 3 */
348 add r6,r3,r5
349 add r4,r4,r5
350 beq 2f
351 andi. r0,r6,3
352 mtctr r7
353 bne 5f
3541: lwz r7,-4(r4)
355 lwzu r8,-8(r4)
356 stw r7,-4(r6)
357 stwu r8,-8(r6)
358 bdnz 1b
359 andi. r5,r5,7
3602: cmplwi 0,r5,4
361 blt 3f
362 lwzu r0,-4(r4)
363 subi r5,r5,4
364 stwu r0,-4(r6)
3653: cmpwi 0,r5,0
366 beqlr
367 mtctr r5
3684: lbzu r0,-1(r4)
369 stbu r0,-1(r6)
370 bdnz 4b
371 blr
3725: mtctr r0
3736: lbzu r7,-1(r4)
374 stbu r7,-1(r6)
375 bdnz 6b
376 subf r5,r0,r5
377 rlwinm. r7,r5,32-3,3,31
378 beq 2b
379 mtctr r7
380 b 1b
381
382_GLOBAL(memcmp)
383 cmpwi 0,r5,0
384 ble- 2f
385 mtctr r5
386 addi r6,r3,-1
387 addi r4,r4,-1
3881: lbzu r3,1(r6)
389 lbzu r0,1(r4)
390 subf. r3,r0,r3
391 bdnzt 2,1b
392 blr
3932: li r3,0
394 blr
395
396_GLOBAL(memchr)
397 cmpwi 0,r5,0
398 ble- 2f
399 mtctr r5
400 addi r3,r3,-1
4011: lbzu r0,1(r3)
402 cmpw 0,r0,r4
403 bdnzf 2,1b
404 beqlr
4052: li r3,0
406 blr
407
408_GLOBAL(__copy_tofrom_user)
409 addi r4,r4,-4
410 addi r6,r3,-4
411 neg r0,r3
412 andi. r0,r0,CACHELINE_MASK /* # bytes to start of cache line */
413 beq 58f
414
415 cmplw 0,r5,r0 /* is this more than total to do? */
416 blt 63f /* if not much to do */
417 andi. r8,r0,3 /* get it word-aligned first */
418 mtctr r8
419 beq+ 61f
42070: lbz r9,4(r4) /* do some bytes */
42171: stb r9,4(r6)
422 addi r4,r4,1
423 addi r6,r6,1
424 bdnz 70b
42561: subf r5,r0,r5
426 srwi. r0,r0,2
427 mtctr r0
428 beq 58f
42972: lwzu r9,4(r4) /* do some words */
43073: stwu r9,4(r6)
431 bdnz 72b
432
433 .section __ex_table,"a"
434 .align 2
435 .long 70b,100f
436 .long 71b,101f
437 .long 72b,102f
438 .long 73b,103f
439 .text
440
44158: srwi. r0,r5,LG_CACHELINE_BYTES /* # complete cachelines */
442 clrlwi r5,r5,32-LG_CACHELINE_BYTES
443 li r11,4
444 beq 63f
445
446#ifdef CONFIG_8xx
447 /* Don't use prefetch on 8xx */
448 mtctr r0
44953: COPY_16_BYTES_WITHEX(0)
450 bdnz 53b
451
452#else /* not CONFIG_8xx */
453 /* Here we decide how far ahead to prefetch the source */
454 li r3,4
455 cmpwi r0,1
456 li r7,0
457 ble 114f
458 li r7,1
459#if MAX_COPY_PREFETCH > 1
460 /* Heuristically, for large transfers we prefetch
461 MAX_COPY_PREFETCH cachelines ahead. For small transfers
462 we prefetch 1 cacheline ahead. */
463 cmpwi r0,MAX_COPY_PREFETCH
464 ble 112f
465 li r7,MAX_COPY_PREFETCH
466112: mtctr r7
467111: dcbt r3,r4
468 addi r3,r3,CACHELINE_BYTES
469 bdnz 111b
470#else
471 dcbt r3,r4
472 addi r3,r3,CACHELINE_BYTES
473#endif /* MAX_COPY_PREFETCH > 1 */
474
475114: subf r8,r7,r0
476 mr r0,r7
477 mtctr r8
478
47953: dcbt r3,r4
48054: dcbz r11,r6
481 .section __ex_table,"a"
482 .align 2
483 .long 54b,105f
484 .text
485/* the main body of the cacheline loop */
486 COPY_16_BYTES_WITHEX(0)
487#if L1_CACHE_LINE_SIZE >= 32
488 COPY_16_BYTES_WITHEX(1)
489#if L1_CACHE_LINE_SIZE >= 64
490 COPY_16_BYTES_WITHEX(2)
491 COPY_16_BYTES_WITHEX(3)
492#if L1_CACHE_LINE_SIZE >= 128
493 COPY_16_BYTES_WITHEX(4)
494 COPY_16_BYTES_WITHEX(5)
495 COPY_16_BYTES_WITHEX(6)
496 COPY_16_BYTES_WITHEX(7)
497#endif
498#endif
499#endif
500 bdnz 53b
501 cmpwi r0,0
502 li r3,4
503 li r7,0
504 bne 114b
505#endif /* CONFIG_8xx */
506
50763: srwi. r0,r5,2
508 mtctr r0
509 beq 64f
51030: lwzu r0,4(r4)
51131: stwu r0,4(r6)
512 bdnz 30b
513
51464: andi. r0,r5,3
515 mtctr r0
516 beq+ 65f
51740: lbz r0,4(r4)
51841: stb r0,4(r6)
519 addi r4,r4,1
520 addi r6,r6,1
521 bdnz 40b
52265: li r3,0
523 blr
524
525/* read fault, initial single-byte copy */
526100: li r9,0
527 b 90f
528/* write fault, initial single-byte copy */
529101: li r9,1
53090: subf r5,r8,r5
531 li r3,0
532 b 99f
533/* read fault, initial word copy */
534102: li r9,0
535 b 91f
536/* write fault, initial word copy */
537103: li r9,1
53891: li r3,2
539 b 99f
540
541/*
542 * this stuff handles faults in the cacheline loop and branches to either
543 * 104f (if in read part) or 105f (if in write part), after updating r5
544 */
545 COPY_16_BYTES_EXCODE(0)
546#if L1_CACHE_LINE_SIZE >= 32
547 COPY_16_BYTES_EXCODE(1)
548#if L1_CACHE_LINE_SIZE >= 64
549 COPY_16_BYTES_EXCODE(2)
550 COPY_16_BYTES_EXCODE(3)
551#if L1_CACHE_LINE_SIZE >= 128
552 COPY_16_BYTES_EXCODE(4)
553 COPY_16_BYTES_EXCODE(5)
554 COPY_16_BYTES_EXCODE(6)
555 COPY_16_BYTES_EXCODE(7)
556#endif
557#endif
558#endif
559
560/* read fault in cacheline loop */
561104: li r9,0
562 b 92f
563/* fault on dcbz (effectively a write fault) */
564/* or write fault in cacheline loop */
565105: li r9,1
56692: li r3,LG_CACHELINE_BYTES
567 b 99f
568/* read fault in final word loop */
569108: li r9,0
570 b 93f
571/* write fault in final word loop */
572109: li r9,1
57393: andi. r5,r5,3
574 li r3,2
575 b 99f
576/* read fault in final byte loop */
577110: li r9,0
578 b 94f
579/* write fault in final byte loop */
580111: li r9,1
58194: li r5,0
582 li r3,0
583/*
584 * At this stage the number of bytes not copied is
585 * r5 + (ctr << r3), and r9 is 0 for read or 1 for write.
586 */
58799: mfctr r0
588 slw r3,r0,r3
589 add. r3,r3,r5
590 beq 120f /* shouldn't happen */
591 cmpwi 0,r9,0
592 bne 120f
593/* for a read fault, first try to continue the copy one byte at a time */
594 mtctr r3
595130: lbz r0,4(r4)
596131: stb r0,4(r6)
597 addi r4,r4,1
598 addi r6,r6,1
599 bdnz 130b
600/* then clear out the destination: r3 bytes starting at 4(r6) */
601132: mfctr r3
602 srwi. r0,r3,2
603 li r9,0
604 mtctr r0
605 beq 113f
606112: stwu r9,4(r6)
607 bdnz 112b
608113: andi. r0,r3,3
609 mtctr r0
610 beq 120f
611114: stb r9,4(r6)
612 addi r6,r6,1
613 bdnz 114b
614120: blr
615
616 .section __ex_table,"a"
617 .align 2
618 .long 30b,108b
619 .long 31b,109b
620 .long 40b,110b
621 .long 41b,111b
622 .long 130b,132b
623 .long 131b,120b
624 .long 112b,120b
625 .long 114b,120b
626 .text
627
628_GLOBAL(__clear_user)
629 addi r6,r3,-4
630 li r3,0
631 li r5,0
632 cmplwi 0,r4,4
633 blt 7f
634 /* clear a single word */
63511: stwu r5,4(r6)
636 beqlr
637 /* clear word sized chunks */
638 andi. r0,r6,3
639 add r4,r0,r4
640 subf r6,r0,r6
641 srwi r0,r4,2
642 andi. r4,r4,3
643 mtctr r0
644 bdz 7f
6451: stwu r5,4(r6)
646 bdnz 1b
647 /* clear byte sized chunks */
6487: cmpwi 0,r4,0
649 beqlr
650 mtctr r4
651 addi r6,r6,3
6528: stbu r5,1(r6)
653 bdnz 8b
654 blr
65590: mr r3,r4
656 blr
65791: mfctr r3
658 slwi r3,r3,2
659 add r3,r3,r4
660 blr
66192: mfctr r3
662 blr
663
664 .section __ex_table,"a"
665 .align 2
666 .long 11b,90b
667 .long 1b,91b
668 .long 8b,92b
669 .text
670
671_GLOBAL(__strncpy_from_user)
672 addi r6,r3,-1
673 addi r4,r4,-1
674 cmpwi 0,r5,0
675 beq 2f
676 mtctr r5
6771: lbzu r0,1(r4)
678 cmpwi 0,r0,0
679 stbu r0,1(r6)
680 bdnzf 2,1b /* dec ctr, branch if ctr != 0 && !cr0.eq */
681 beq 3f
6822: addi r6,r6,1
6833: subf r3,r3,r6
684 blr
68599: li r3,-EFAULT
686 blr
687
688 .section __ex_table,"a"
689 .align 2
690 .long 1b,99b
691 .text
692
693/* r3 = str, r4 = len (> 0), r5 = top (highest addr) */
694_GLOBAL(__strnlen_user)
695 addi r7,r3,-1
696 subf r6,r7,r5 /* top+1 - str */
697 cmplw 0,r4,r6
698 bge 0f
699 mr r6,r4
7000: mtctr r6 /* ctr = min(len, top - str) */
7011: lbzu r0,1(r7) /* get next byte */
702 cmpwi 0,r0,0
703 bdnzf 2,1b /* loop if --ctr != 0 && byte != 0 */
704 addi r7,r7,1
705 subf r3,r3,r7 /* number of bytes we have looked at */
706 beqlr /* return if we found a 0 byte */
707 cmpw 0,r3,r4 /* did we look at all len bytes? */
708 blt 99f /* if not, must have hit top */
709 addi r3,r4,1 /* return len + 1 to indicate no null found */
710 blr
71199: li r3,0 /* bad address, return 0 */
712 blr
713
714 .section __ex_table,"a"
715 .align 2
716 .long 1b,99b
generated by cgit v1.2.2 (git 2.25.0) at 2024-10-06 20:21:12 -0400