forked from ~ljy/RK356X_SDK_RELEASE

blame | history | patch | commit | commitdiff
hc
2024-09-20 a36159eec6ca17402b0e146b86efaf76568dc353 
 kernel/arch/x86/lib/checksum_32.S
....@@ -1,3 +1,4 @@
1
+/* SPDX-License-Identifier: GPL-2.0-or-later */
12 /*
23  * INET		An implementation of the TCP/IP protocol suite for the LINUX
34  *		operating system.  INET is implemented using the  BSD Socket
....@@ -18,11 +19,6 @@
1819  *			     handling.
1920  *		Andi Kleen,  add zeroing on error
2021  *                   converted to pure assembler
21
- *
22
- *		This program is free software; you can redistribute it and/or
23
- *		modify it under the terms of the GNU General Public License
24
- *		as published by the Free Software Foundation; either version
25
- *		2 of the License, or (at your option) any later version.
2622  */
2723 
2824 #include <linux/linkage.h>
....@@ -50,7 +46,7 @@
5046 	   * Fortunately, it is easy to convert 2-byte alignment to 4-byte
5147 	   * alignment for the unrolled loop.
5248 	   */		
53
-ENTRY(csum_partial)
49
+SYM_FUNC_START(csum_partial)
5450 	pushl %esi
5551 	pushl %ebx
5652 	movl 20(%esp),%eax	# Function arg: unsigned int sum
....@@ -131,14 +127,14 @@
131127 8:
132128 	popl %ebx
133129 	popl %esi
134
-	ret
135
-ENDPROC(csum_partial)
130
+	RET
131
+SYM_FUNC_END(csum_partial)
136132 
137133 #else
138134 
139135 /* Version for PentiumII/PPro */
140136 
141
-ENTRY(csum_partial)
137
+SYM_FUNC_START(csum_partial)
142138 	pushl %esi
143139 	pushl %ebx
144140 	movl 20(%esp),%eax	# Function arg: unsigned int sum
....@@ -157,7 +153,7 @@
157153 	negl %ebx
158154 	lea 45f(%ebx,%ebx,2), %ebx
159155 	testl %esi, %esi
160
-	JMP_NOSPEC %ebx
156
+	JMP_NOSPEC ebx
161157 
162158 	# Handle 2-byte-aligned regions
163159 20:	addw (%esi), %ax
....@@ -249,60 +245,49 @@
249245 90: 
250246 	popl %ebx
251247 	popl %esi
252
-	ret
253
-ENDPROC(csum_partial)
248
+	RET
249
+SYM_FUNC_END(csum_partial)
254250 				
255251 #endif
256252 EXPORT_SYMBOL(csum_partial)
257253 
258254 /*
259255 unsigned int csum_partial_copy_generic (const char *src, char *dst,
260
-				  int len, int sum, int *src_err_ptr, int *dst_err_ptr)
256
+				  int len)
261257  */ 
262258 
263259 /*
264260  * Copy from ds while checksumming, otherwise like csum_partial
265
- *
266
- * The macros SRC and DST specify the type of access for the instruction.
267
- * thus we can call a custom exception handler for all access types.
268
- *
269
- * FIXME: could someone double-check whether I haven't mixed up some SRC and
270
- *	  DST definitions? It's damn hard to trigger all cases.  I hope I got
271
- *	  them all but there's no guarantee.
272261  */
273262 
274
-#define SRC(y...)			\
263
+#define EXC(y...)			\
275264 	9999: y;			\
276
-	_ASM_EXTABLE(9999b, 6001f)
277
-
278
-#define DST(y...)			\
279
-	9999: y;			\
280
-	_ASM_EXTABLE(9999b, 6002f)
265
+	_ASM_EXTABLE_UA(9999b, 6001f)
281266 
282267 #ifndef CONFIG_X86_USE_PPRO_CHECKSUM
283268 
284269 #define ARGBASE 16		
285270 #define FP		12
286271 		
287
-ENTRY(csum_partial_copy_generic)
272
+SYM_FUNC_START(csum_partial_copy_generic)
288273 	subl  $4,%esp	
289274 	pushl %edi
290275 	pushl %esi
291276 	pushl %ebx
292
-	movl ARGBASE+16(%esp),%eax	# sum
293277 	movl ARGBASE+12(%esp),%ecx	# len
294278 	movl ARGBASE+4(%esp),%esi	# src
295279 	movl ARGBASE+8(%esp),%edi	# dst
296280 
281
+	movl $-1, %eax			# sum
297282 	testl $2, %edi			# Check alignment. 
298283 	jz 2f				# Jump if alignment is ok.
299284 	subl $2, %ecx			# Alignment uses up two bytes.
300285 	jae 1f				# Jump if we had at least two bytes.
301286 	addl $2, %ecx			# ecx was < 2.  Deal with it.
302287 	jmp 4f
303
-SRC(1:	movw (%esi), %bx	)
288
+EXC(1:	movw (%esi), %bx	)
304289 	addl $2, %esi
305
-DST(	movw %bx, (%edi)	)
290
+EXC(	movw %bx, (%edi)	)
306291 	addl $2, %edi
307292 	addw %bx, %ax	
308293 	adcl $0, %eax
....@@ -310,34 +295,34 @@
310295 	movl %ecx, FP(%esp)
311296 	shrl $5, %ecx
312297 	jz 2f
313
-	testl %esi, %esi
314
-SRC(1:	movl (%esi), %ebx	)
315
-SRC(	movl 4(%esi), %edx	)
298
+	testl %esi, %esi		# what's wrong with clc?
299
+EXC(1:	movl (%esi), %ebx	)
300
+EXC(	movl 4(%esi), %edx	)
316301 	adcl %ebx, %eax
317
-DST(	movl %ebx, (%edi)	)
302
+EXC(	movl %ebx, (%edi)	)
318303 	adcl %edx, %eax
319
-DST(	movl %edx, 4(%edi)	)
304
+EXC(	movl %edx, 4(%edi)	)
320305 
321
-SRC(	movl 8(%esi), %ebx	)
322
-SRC(	movl 12(%esi), %edx	)
306
+EXC(	movl 8(%esi), %ebx	)
307
+EXC(	movl 12(%esi), %edx	)
323308 	adcl %ebx, %eax
324
-DST(	movl %ebx, 8(%edi)	)
309
+EXC(	movl %ebx, 8(%edi)	)
325310 	adcl %edx, %eax
326
-DST(	movl %edx, 12(%edi)	)
311
+EXC(	movl %edx, 12(%edi)	)
327312 
328
-SRC(	movl 16(%esi), %ebx 	)
329
-SRC(	movl 20(%esi), %edx	)
313
+EXC(	movl 16(%esi), %ebx 	)
314
+EXC(	movl 20(%esi), %edx	)
330315 	adcl %ebx, %eax
331
-DST(	movl %ebx, 16(%edi)	)
316
+EXC(	movl %ebx, 16(%edi)	)
332317 	adcl %edx, %eax
333
-DST(	movl %edx, 20(%edi)	)
318
+EXC(	movl %edx, 20(%edi)	)
334319 
335
-SRC(	movl 24(%esi), %ebx	)
336
-SRC(	movl 28(%esi), %edx	)
320
+EXC(	movl 24(%esi), %ebx	)
321
+EXC(	movl 28(%esi), %edx	)
337322 	adcl %ebx, %eax
338
-DST(	movl %ebx, 24(%edi)	)
323
+EXC(	movl %ebx, 24(%edi)	)
339324 	adcl %edx, %eax
340
-DST(	movl %edx, 28(%edi)	)
325
+EXC(	movl %edx, 28(%edi)	)
341326 
342327 	lea 32(%esi), %esi
343328 	lea 32(%edi), %edi
....@@ -349,9 +334,9 @@
349334 	andl $0x1c, %edx
350335 	je 4f
351336 	shrl $2, %edx			# This clears CF
352
-SRC(3:	movl (%esi), %ebx	)
337
+EXC(3:	movl (%esi), %ebx	)
353338 	adcl %ebx, %eax
354
-DST(	movl %ebx, (%edi)	)
339
+EXC(	movl %ebx, (%edi)	)
355340 	lea 4(%esi), %esi
356341 	lea 4(%edi), %edi
357342 	dec %edx
....@@ -361,39 +346,24 @@
361346 	jz 7f
362347 	cmpl $2, %ecx
363348 	jb 5f
364
-SRC(	movw (%esi), %cx	)
349
+EXC(	movw (%esi), %cx	)
365350 	leal 2(%esi), %esi
366
-DST(	movw %cx, (%edi)	)
351
+EXC(	movw %cx, (%edi)	)
367352 	leal 2(%edi), %edi
368353 	je 6f
369354 	shll $16,%ecx
370
-SRC(5:	movb (%esi), %cl	)
371
-DST(	movb %cl, (%edi)	)
355
+EXC(5:	movb (%esi), %cl	)
356
+EXC(	movb %cl, (%edi)	)
372357 6:	addl %ecx, %eax
373358 	adcl $0, %eax
374359 7:
375
-5000:
376360 
377361 # Exception handler:
378362 .section .fixup, "ax"							
379363 
380364 6001:
381
-	movl ARGBASE+20(%esp), %ebx	# src_err_ptr
382
-	movl $-EFAULT, (%ebx)
383
-
384
-	# zero the complete destination - computing the rest
385
-	# is too much work 
386
-	movl ARGBASE+8(%esp), %edi	# dst
387
-	movl ARGBASE+12(%esp), %ecx	# len
388
-	xorl %eax,%eax
389
-	rep ; stosb
390
-
391
-	jmp 5000b
392
-
393
-6002:
394
-	movl ARGBASE+24(%esp), %ebx	# dst_err_ptr
395
-	movl $-EFAULT,(%ebx)
396
-	jmp 5000b
365
+	xorl %eax, %eax
366
+	jmp 7b
397367 
398368 .previous
399369 
....@@ -401,33 +371,33 @@
401371 	popl %esi
402372 	popl %edi
403373 	popl %ecx			# equivalent to addl $4,%esp
404
-	ret	
405
-ENDPROC(csum_partial_copy_generic)
374
+	RET
375
+SYM_FUNC_END(csum_partial_copy_generic)
406376 
407377 #else
408378 
409379 /* Version for PentiumII/PPro */
410380 
411381 #define ROUND1(x) \
412
-	SRC(movl x(%esi), %ebx	)	;	\
382
+	EXC(movl x(%esi), %ebx	)	;	\
413383 	addl %ebx, %eax			;	\
414
-	DST(movl %ebx, x(%edi)	)	; 
384
+	EXC(movl %ebx, x(%edi)	)	;
415385 
416386 #define ROUND(x) \
417
-	SRC(movl x(%esi), %ebx	)	;	\
387
+	EXC(movl x(%esi), %ebx	)	;	\
418388 	adcl %ebx, %eax			;	\
419
-	DST(movl %ebx, x(%edi)	)	;
389
+	EXC(movl %ebx, x(%edi)	)	;
420390 
421391 #define ARGBASE 12
422392 		
423
-ENTRY(csum_partial_copy_generic)
393
+SYM_FUNC_START(csum_partial_copy_generic)
424394 	pushl %ebx
425395 	pushl %edi
426396 	pushl %esi
427397 	movl ARGBASE+4(%esp),%esi	#src
428398 	movl ARGBASE+8(%esp),%edi	#dst	
429399 	movl ARGBASE+12(%esp),%ecx	#len
430
-	movl ARGBASE+16(%esp),%eax	#sum
400
+	movl $-1, %eax			#sum
431401 #	movl %ecx, %edx  
432402 	movl %ecx, %ebx  
433403 	movl %esi, %edx
....@@ -440,10 +410,10 @@
440410 	andl $-32,%edx
441411 	lea 3f(%ebx,%ebx), %ebx
442412 	testl %esi, %esi 
443
-	JMP_NOSPEC %ebx
413
+	JMP_NOSPEC ebx
444414 1:	addl $64,%esi
445415 	addl $64,%edi 
446
-	SRC(movb -32(%edx),%bl)	; SRC(movb (%edx),%bl)
416
+	EXC(movb -32(%edx),%bl)	; EXC(movb (%edx),%bl)
447417 	ROUND1(-64) ROUND(-60) ROUND(-56) ROUND(-52)	
448418 	ROUND (-48) ROUND(-44) ROUND(-40) ROUND(-36)	
449419 	ROUND (-32) ROUND(-28) ROUND(-24) ROUND(-20)	
....@@ -457,37 +427,28 @@
457427 	jz 7f
458428 	cmpl $2, %edx
459429 	jb 5f
460
-SRC(	movw (%esi), %dx         )
430
+EXC(	movw (%esi), %dx         )
461431 	leal 2(%esi), %esi
462
-DST(	movw %dx, (%edi)         )
432
+EXC(	movw %dx, (%edi)         )
463433 	leal 2(%edi), %edi
464434 	je 6f
465435 	shll $16,%edx
466436 5:
467
-SRC(	movb (%esi), %dl         )
468
-DST(	movb %dl, (%edi)         )
437
+EXC(	movb (%esi), %dl         )
438
+EXC(	movb %dl, (%edi)         )
469439 6:	addl %edx, %eax
470440 	adcl $0, %eax
471441 7:
472442 .section .fixup, "ax"
473
-6001:	movl	ARGBASE+20(%esp), %ebx	# src_err_ptr	
474
-	movl $-EFAULT, (%ebx)
475
-	# zero the complete destination (computing the rest is too much work)
476
-	movl ARGBASE+8(%esp),%edi	# dst
477
-	movl ARGBASE+12(%esp),%ecx	# len
478
-	xorl %eax,%eax
479
-	rep; stosb
480
-	jmp 7b
481
-6002:	movl ARGBASE+24(%esp), %ebx	# dst_err_ptr
482
-	movl $-EFAULT, (%ebx)
443
+6001:	xorl %eax, %eax
483444 	jmp  7b			
484445 .previous				
485446 
486447 	popl %esi
487448 	popl %edi
488449 	popl %ebx
489
-	ret
490
-ENDPROC(csum_partial_copy_generic)
450
+	RET
451
+SYM_FUNC_END(csum_partial_copy_generic)
491452 				
492453 #undef ROUND
493454 #undef ROUND1