crypto/fsl: Fix HW accelerated hash commands

The hash command function were not flushing the dcache before passing data to CAAM/DMA and not invalidating the dcache when getting data back. Due the data cache incoherency, HW accelerated hash commands used to fail with CAAM errors like "Invalid KEY Command". Check if pbuf and pout buffers are properly aligned to the cache line size and flush/invalidate the memory regions to address this issue. This solution is based in a previous work from Clemens Gruber in commit 598e9dccc75d ("crypto/fsl: fix BLOB encapsulation and decapsulation") Reported-by: Anatolij Gustschin <agust@denx.de> Signed-off-by: Breno Lima <breno.lima@nxp.com> Reviewed-by: York Sun <york.sun@nxp.com>

crypto/fsl: Fix HW accelerated hash commands
The hash command function were not flushing the dcache before passing data to CAAM/DMA and not invalidating the dcache when getting data back. Due the data cache incoherency, HW accelerated hash commands used to fail with CAAM errors like "Invalid KEY Command". Check if pbuf and pout buffers are properly aligned to the cache line size and flush/invalidate the memory regions to address this issue. This solution is based in a previous work from Clemens Gruber in commit 598e9dccc75d ("crypto/fsl: fix BLOB encapsulation and decapsulation") Reported-by: Anatolij Gustschin <agust@denx.de> Signed-off-by: Breno Lima <breno.lima@nxp.com> Reviewed-by: York Sun <york.sun@nxp.com>
Breno Lima · York Sun · Eric Lee
1 parent 6d48d1c4b4
Showing 2 changed files with 25 additions and 2 deletions Side-by-side Diff
common/hash.c
drivers/crypto/fsl/fsl_hash.c
@@ -390,7 +390,7 @@
  
 	if (multi_hash()) {
 		struct hash_algo *algo;
-		uint8_t output[HASH_MAX_DIGEST_SIZE];
+		u8 *output;
 		uint8_t vsum[HASH_MAX_DIGEST_SIZE];
 		void *buf;
  
@@ -405,6 +405,9 @@
 			return 1;
 		}
  
+		output = memalign(ARCH_DMA_MINALIGN,
+				  sizeof(uint32_t) * HASH_MAX_DIGEST_SIZE);
+
 		buf = map_sysmem(addr, len);
 		algo->hash_func_ws(buf, len, output, algo->chunk_size);
 		unmap_sysmem(buf);
@@ -440,6 +443,8 @@
 				store_result(algo, output, *argv,
 					flags & HASH_FLAG_ENV);
 			}
+		unmap_sysmem(output);
+
 		}
  
 	/* Horrible code size hack for boards that just want crc32 */
@@ -7,6 +7,7 @@
  
 #include <common.h>
 #include <malloc.h>
+#include <memalign.h>
 #include "jobdesc.h"
 #include "desc.h"
 #include "jr.h"
  
  
  
  
@@ -163,19 +164,36 @@
 {
 	int ret = 0;
 	uint32_t *desc;
+	unsigned int size;
  
-	desc = malloc(sizeof(int) * MAX_CAAM_DESCSIZE);
+	desc = malloc_cache_aligned(sizeof(int) * MAX_CAAM_DESCSIZE);
 	if (!desc) {
 		debug("Not enough memory for descriptor allocation\n");
 		return -ENOMEM;
 	}
  
+	if (!IS_ALIGNED((uintptr_t)pbuf, ARCH_DMA_MINALIGN) ||
+	    !IS_ALIGNED((uintptr_t)pout, ARCH_DMA_MINALIGN)) {
+		puts("Error: Address arguments are not aligned\n");
+		return -EINVAL;
+	}
+
+	size = ALIGN(buf_len, ARCH_DMA_MINALIGN);
+	flush_dcache_range((unsigned long)pbuf, (unsigned long)pbuf + size);
+
 	inline_cnstr_jobdesc_hash(desc, pbuf, buf_len, pout,
 				  driver_hash[algo].alg_type,
 				  driver_hash[algo].digestsize,
 				  0);
  
+	size = ALIGN(sizeof(int) * MAX_CAAM_DESCSIZE, ARCH_DMA_MINALIGN);
+	flush_dcache_range((unsigned long)desc, (unsigned long)desc + size);
+
 	ret = run_descriptor_jr(desc);
+
+	size = ALIGN(driver_hash[algo].digestsize, ARCH_DMA_MINALIGN);
+	invalidate_dcache_range((unsigned long)pout,
+				(unsigned long)pout + size);
  
 	free(desc);
 	return ret;
...	...	@@ -390,7 +390,7 @@
390	390
391	391	if (multi_hash()) {
392	392	struct hash_algo *algo;
393		- uint8_t output[HASH_MAX_DIGEST_SIZE];
	393	+ u8 *output;
394	394	uint8_t vsum[HASH_MAX_DIGEST_SIZE];
395	395	void *buf;
396	396
...	...	@@ -405,6 +405,9 @@
405	405	return 1;
406	406	}
407	407
	408	+ output = memalign(ARCH_DMA_MINALIGN,
	409	+ sizeof(uint32_t) * HASH_MAX_DIGEST_SIZE);
	410	+
408	411	buf = map_sysmem(addr, len);
409	412	algo->hash_func_ws(buf, len, output, algo->chunk_size);
410	413	unmap_sysmem(buf);
...	...	@@ -440,6 +443,8 @@
440	443	store_result(algo, output, *argv,
441	444	flags & HASH_FLAG_ENV);
442	445	}
	446	+ unmap_sysmem(output);
	447	+
443	448	}
444	449
445	450	/* Horrible code size hack for boards that just want crc32 */
...	...	@@ -7,6 +7,7 @@
7	7
8	8	#include <common.h>
9	9	#include <malloc.h>
	10	+#include <memalign.h>
10	11	#include "jobdesc.h"
11	12	#include "desc.h"
12	13	#include "jr.h"
13	14
14	15
15	16
16	17
...	...	@@ -163,19 +164,36 @@
163	164	{
164	165	int ret = 0;
165	166	uint32_t *desc;
	167	+ unsigned int size;
166	168
167		- desc = malloc(sizeof(int) * MAX_CAAM_DESCSIZE);
	169	+ desc = malloc_cache_aligned(sizeof(int) * MAX_CAAM_DESCSIZE);
168	170	if (!desc) {
169	171	debug("Not enough memory for descriptor allocation\n");
170	172	return -ENOMEM;
171	173	}
172	174
	175	+ if (!IS_ALIGNED((uintptr_t)pbuf, ARCH_DMA_MINALIGN) \|\|
	176	+ !IS_ALIGNED((uintptr_t)pout, ARCH_DMA_MINALIGN)) {
	177	+ puts("Error: Address arguments are not aligned\n");
	178	+ return -EINVAL;
	179	+ }
	180	+
	181	+ size = ALIGN(buf_len, ARCH_DMA_MINALIGN);
	182	+ flush_dcache_range((unsigned long)pbuf, (unsigned long)pbuf + size);
	183	+
173	184	inline_cnstr_jobdesc_hash(desc, pbuf, buf_len, pout,
174	185	driver_hash[algo].alg_type,
175	186	driver_hash[algo].digestsize,
176	187	0);
177	188
	189	+ size = ALIGN(sizeof(int) * MAX_CAAM_DESCSIZE, ARCH_DMA_MINALIGN);
	190	+ flush_dcache_range((unsigned long)desc, (unsigned long)desc + size);
	191	+
178	192	ret = run_descriptor_jr(desc);
	193	+
	194	+ size = ALIGN(driver_hash[algo].digestsize, ARCH_DMA_MINALIGN);
	195	+ invalidate_dcache_range((unsigned long)pout,
	196	+ (unsigned long)pout + size);
179	197
180	198	free(desc);
181	199	return ret;