1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3 * AMD Cryptographic Coprocessor (CCP) driver
4 *
5 * Copyright (C) 2013,2017 Advanced Micro Devices, Inc.
6 *
7 * Author: Tom Lendacky <thomas.lendacky@amd.com>
8 * Author: Gary R Hook <gary.hook@amd.com>
9 */
10
11 #ifndef __CCP_DEV_H__
12 #define __CCP_DEV_H__
13
14 #include <linux/device.h>
15 #include <linux/spinlock.h>
16 #include <linux/mutex.h>
17 #include <linux/list.h>
18 #include <linux/wait.h>
19 #include <linux/dma-direction.h>
20 #include <linux/dmapool.h>
21 #include <linux/hw_random.h>
22 #include <linux/bitops.h>
23 #include <linux/interrupt.h>
24 #include <linux/irqreturn.h>
25 #include <linux/dmaengine.h>
26
27 #include "sp-dev.h"
28
29 #define MAX_CCP_NAME_LEN 16
30 #define MAX_DMAPOOL_NAME_LEN 32
31
32 #define MAX_HW_QUEUES 5
33 #define MAX_CMD_QLEN 100
34
35 #define TRNG_RETRIES 10
36
37 #define CACHE_NONE 0x00
38 #define CACHE_WB_NO_ALLOC 0xb7
39
40 /****** Register Mappings ******/
41 #define Q_MASK_REG 0x000
42 #define TRNG_OUT_REG 0x00c
43 #define IRQ_MASK_REG 0x040
44 #define IRQ_STATUS_REG 0x200
45
46 #define DEL_CMD_Q_JOB 0x124
47 #define DEL_Q_ACTIVE 0x00000200
48 #define DEL_Q_ID_SHIFT 6
49
50 #define CMD_REQ0 0x180
51 #define CMD_REQ_INCR 0x04
52
53 #define CMD_Q_STATUS_BASE 0x210
54 #define CMD_Q_INT_STATUS_BASE 0x214
55 #define CMD_Q_STATUS_INCR 0x20
56
57 #define CMD_Q_CACHE_BASE 0x228
58 #define CMD_Q_CACHE_INC 0x20
59
60 #define CMD_Q_ERROR(__qs) ((__qs) & 0x0000003f)
61 #define CMD_Q_DEPTH(__qs) (((__qs) >> 12) & 0x0000000f)
62
63 /* ------------------------ CCP Version 5 Specifics ------------------------ */
64 #define CMD5_QUEUE_MASK_OFFSET 0x00
65 #define CMD5_QUEUE_PRIO_OFFSET 0x04
66 #define CMD5_REQID_CONFIG_OFFSET 0x08
67 #define CMD5_CMD_TIMEOUT_OFFSET 0x10
68 #define LSB_PUBLIC_MASK_LO_OFFSET 0x18
69 #define LSB_PUBLIC_MASK_HI_OFFSET 0x1C
70 #define LSB_PRIVATE_MASK_LO_OFFSET 0x20
71 #define LSB_PRIVATE_MASK_HI_OFFSET 0x24
72 #define CMD5_PSP_CCP_VERSION 0x100
73
74 #define CMD5_Q_CONTROL_BASE 0x0000
75 #define CMD5_Q_TAIL_LO_BASE 0x0004
76 #define CMD5_Q_HEAD_LO_BASE 0x0008
77 #define CMD5_Q_INT_ENABLE_BASE 0x000C
78 #define CMD5_Q_INTERRUPT_STATUS_BASE 0x0010
79
80 #define CMD5_Q_STATUS_BASE 0x0100
81 #define CMD5_Q_INT_STATUS_BASE 0x0104
82 #define CMD5_Q_DMA_STATUS_BASE 0x0108
83 #define CMD5_Q_DMA_READ_STATUS_BASE 0x010C
84 #define CMD5_Q_DMA_WRITE_STATUS_BASE 0x0110
85 #define CMD5_Q_ABORT_BASE 0x0114
86 #define CMD5_Q_AX_CACHE_BASE 0x0118
87
88 #define CMD5_CONFIG_0_OFFSET 0x6000
89 #define CMD5_TRNG_CTL_OFFSET 0x6008
90 #define CMD5_AES_MASK_OFFSET 0x6010
91 #define CMD5_CLK_GATE_CTL_OFFSET 0x603C
92
93 /* Address offset between two virtual queue registers */
94 #define CMD5_Q_STATUS_INCR 0x1000
95
96 /* Bit masks */
97 #define CMD5_Q_RUN 0x1
98 #define CMD5_Q_HALT 0x2
99 #define CMD5_Q_MEM_LOCATION 0x4
100 #define CMD5_Q_SIZE 0x1F
101 #define CMD5_Q_SHIFT 3
102 #define COMMANDS_PER_QUEUE 16
103 #define QUEUE_SIZE_VAL ((ffs(COMMANDS_PER_QUEUE) - 2) & \
104 CMD5_Q_SIZE)
105 #define Q_PTR_MASK (2 << (QUEUE_SIZE_VAL + 5) - 1)
106 #define Q_DESC_SIZE sizeof(struct ccp5_desc)
107 #define Q_SIZE(n) (COMMANDS_PER_QUEUE*(n))
108
109 #define INT_COMPLETION 0x1
110 #define INT_ERROR 0x2
111 #define INT_QUEUE_STOPPED 0x4
112 #define INT_EMPTY_QUEUE 0x8
113 #define SUPPORTED_INTERRUPTS (INT_COMPLETION | INT_ERROR)
114
115 #define LSB_REGION_WIDTH 5
116 #define MAX_LSB_CNT 8
117
118 #define LSB_SIZE 16
119 #define LSB_ITEM_SIZE 32
120 #define PLSB_MAP_SIZE (LSB_SIZE)
121 #define SLSB_MAP_SIZE (MAX_LSB_CNT * LSB_SIZE)
122
123 #define LSB_ENTRY_NUMBER(LSB_ADDR) (LSB_ADDR / LSB_ITEM_SIZE)
124
125 /* ------------------------ CCP Version 3 Specifics ------------------------ */
126 #define REQ0_WAIT_FOR_WRITE 0x00000004
127 #define REQ0_INT_ON_COMPLETE 0x00000002
128 #define REQ0_STOP_ON_COMPLETE 0x00000001
129
130 #define REQ0_CMD_Q_SHIFT 9
131 #define REQ0_JOBID_SHIFT 3
132
133 /****** REQ1 Related Values ******/
134 #define REQ1_PROTECT_SHIFT 27
135 #define REQ1_ENGINE_SHIFT 23
136 #define REQ1_KEY_KSB_SHIFT 2
137
138 #define REQ1_EOM 0x00000002
139 #define REQ1_INIT 0x00000001
140
141 /* AES Related Values */
142 #define REQ1_AES_TYPE_SHIFT 21
143 #define REQ1_AES_MODE_SHIFT 18
144 #define REQ1_AES_ACTION_SHIFT 17
145 #define REQ1_AES_CFB_SIZE_SHIFT 10
146
147 /* XTS-AES Related Values */
148 #define REQ1_XTS_AES_SIZE_SHIFT 10
149
150 /* SHA Related Values */
151 #define REQ1_SHA_TYPE_SHIFT 21
152
153 /* RSA Related Values */
154 #define REQ1_RSA_MOD_SIZE_SHIFT 10
155
156 /* Pass-Through Related Values */
157 #define REQ1_PT_BW_SHIFT 12
158 #define REQ1_PT_BS_SHIFT 10
159
160 /* ECC Related Values */
161 #define REQ1_ECC_AFFINE_CONVERT 0x00200000
162 #define REQ1_ECC_FUNCTION_SHIFT 18
163
164 /****** REQ4 Related Values ******/
165 #define REQ4_KSB_SHIFT 18
166 #define REQ4_MEMTYPE_SHIFT 16
167
168 /****** REQ6 Related Values ******/
169 #define REQ6_MEMTYPE_SHIFT 16
170
171 /****** Key Storage Block ******/
172 #define KSB_START 77
173 #define KSB_END 127
174 #define KSB_COUNT (KSB_END - KSB_START + 1)
175 #define CCP_SB_BITS 256
176
177 #define CCP_JOBID_MASK 0x0000003f
178
179 /* ------------------------ General CCP Defines ------------------------ */
180
181 #define CCP_DMA_DFLT 0x0
182 #define CCP_DMA_PRIV 0x1
183 #define CCP_DMA_PUB 0x2
184
185 #define CCP_DMAPOOL_MAX_SIZE 64
186 #define CCP_DMAPOOL_ALIGN BIT(5)
187
188 #define CCP_REVERSE_BUF_SIZE 64
189
190 #define CCP_AES_KEY_SB_COUNT 1
191 #define CCP_AES_CTX_SB_COUNT 1
192
193 #define CCP_XTS_AES_KEY_SB_COUNT 1
194 #define CCP5_XTS_AES_KEY_SB_COUNT 2
195 #define CCP_XTS_AES_CTX_SB_COUNT 1
196
197 #define CCP_DES3_KEY_SB_COUNT 1
198 #define CCP_DES3_CTX_SB_COUNT 1
199
200 #define CCP_SHA_SB_COUNT 1
201
202 #define CCP_RSA_MAX_WIDTH 4096
203 #define CCP5_RSA_MAX_WIDTH 16384
204
205 #define CCP_PASSTHRU_BLOCKSIZE 256
206 #define CCP_PASSTHRU_MASKSIZE 32
207 #define CCP_PASSTHRU_SB_COUNT 1
208
209 #define CCP_ECC_MODULUS_BYTES 48 /* 384-bits */
210 #define CCP_ECC_MAX_OPERANDS 6
211 #define CCP_ECC_MAX_OUTPUTS 3
212 #define CCP_ECC_SRC_BUF_SIZE 448
213 #define CCP_ECC_DST_BUF_SIZE 192
214 #define CCP_ECC_OPERAND_SIZE 64
215 #define CCP_ECC_OUTPUT_SIZE 64
216 #define CCP_ECC_RESULT_OFFSET 60
217 #define CCP_ECC_RESULT_SUCCESS 0x0001
218
219 #define CCP_SB_BYTES 32
220
221 struct ccp_op;
222 struct ccp_device;
223 struct ccp_cmd;
224 struct ccp_fns;
225
226 struct ccp_dma_cmd {
227 struct list_head entry;
228
229 struct ccp_cmd ccp_cmd;
230 };
231
232 struct ccp_dma_desc {
233 struct list_head entry;
234
235 struct ccp_device *ccp;
236
237 struct list_head pending;
238 struct list_head active;
239
240 enum dma_status status;
241 struct dma_async_tx_descriptor tx_desc;
242 size_t len;
243 };
244
245 struct ccp_dma_chan {
246 struct ccp_device *ccp;
247
248 spinlock_t lock;
249 struct list_head created;
250 struct list_head pending;
251 struct list_head active;
252 struct list_head complete;
253
254 struct tasklet_struct cleanup_tasklet;
255
256 enum dma_status status;
257 struct dma_chan dma_chan;
258 };
259
260 struct ccp_cmd_queue {
261 struct ccp_device *ccp;
262
263 /* Queue identifier */
264 u32 id;
265
266 /* Queue dma pool */
267 struct dma_pool *dma_pool;
268
269 /* Queue base address (not neccessarily aligned)*/
270 struct ccp5_desc *qbase;
271
272 /* Aligned queue start address (per requirement) */
273 struct mutex q_mutex ____cacheline_aligned;
274 unsigned int qidx;
275
276 /* Version 5 has different requirements for queue memory */
277 unsigned int qsize;
278 dma_addr_t qbase_dma;
279 dma_addr_t qdma_tail;
280
281 /* Per-queue reserved storage block(s) */
282 u32 sb_key;
283 u32 sb_ctx;
284
285 /* Bitmap of LSBs that can be accessed by this queue */
286 DECLARE_BITMAP(lsbmask, MAX_LSB_CNT);
287 /* Private LSB that is assigned to this queue, or -1 if none.
288 * Bitmap for my private LSB, unused otherwise
289 */
290 int lsb;
291 DECLARE_BITMAP(lsbmap, PLSB_MAP_SIZE);
292
293 /* Queue processing thread */
294 struct task_struct *kthread;
295 unsigned int active;
296 unsigned int suspended;
297
298 /* Number of free command slots available */
299 unsigned int free_slots;
300
301 /* Interrupt masks */
302 u32 int_ok;
303 u32 int_err;
304
305 /* Register addresses for queue */
306 void __iomem *reg_control;
307 void __iomem *reg_tail_lo;
308 void __iomem *reg_head_lo;
309 void __iomem *reg_int_enable;
310 void __iomem *reg_interrupt_status;
311 void __iomem *reg_status;
312 void __iomem *reg_int_status;
313 void __iomem *reg_dma_status;
314 void __iomem *reg_dma_read_status;
315 void __iomem *reg_dma_write_status;
316 u32 qcontrol; /* Cached control register */
317
318 /* Status values from job */
319 u32 int_status;
320 u32 q_status;
321 u32 q_int_status;
322 u32 cmd_error;
323
324 /* Interrupt wait queue */
325 wait_queue_head_t int_queue;
326 unsigned int int_rcvd;
327
328 /* Per-queue Statistics */
329 unsigned long total_ops;
330 unsigned long total_aes_ops;
331 unsigned long total_xts_aes_ops;
332 unsigned long total_3des_ops;
333 unsigned long total_sha_ops;
334 unsigned long total_rsa_ops;
335 unsigned long total_pt_ops;
336 unsigned long total_ecc_ops;
337 } ____cacheline_aligned;
338
339 struct ccp_device {
340 struct list_head entry;
341
342 struct ccp_vdata *vdata;
343 unsigned int ord;
344 char name[MAX_CCP_NAME_LEN];
345 char rngname[MAX_CCP_NAME_LEN];
346
347 struct device *dev;
348 struct sp_device *sp;
349
350 /* Bus specific device information
351 */
352 void *dev_specific;
353 unsigned int qim;
354 unsigned int irq;
355 bool use_tasklet;
356 struct tasklet_struct irq_tasklet;
357
358 /* I/O area used for device communication. The register mapping
359 * starts at an offset into the mapped bar.
360 * The CMD_REQx registers and the Delete_Cmd_Queue_Job register
361 * need to be protected while a command queue thread is accessing
362 * them.
363 */
364 struct mutex req_mutex ____cacheline_aligned;
365 void __iomem *io_regs;
366
367 /* Master lists that all cmds are queued on. Because there can be
368 * more than one CCP command queue that can process a cmd a separate
369 * backlog list is neeeded so that the backlog completion call
370 * completes before the cmd is available for execution.
371 */
372 spinlock_t cmd_lock ____cacheline_aligned;
373 unsigned int cmd_count;
374 struct list_head cmd;
375 struct list_head backlog;
376
377 /* The command queues. These represent the queues available on the
378 * CCP that are available for processing cmds
379 */
380 struct ccp_cmd_queue cmd_q[MAX_HW_QUEUES];
381 unsigned int cmd_q_count;
382 unsigned int max_q_count;
383
384 /* Support for the CCP True RNG
385 */
386 struct hwrng hwrng;
387 unsigned int hwrng_retries;
388
389 /* Support for the CCP DMA capabilities
390 */
391 struct dma_device dma_dev;
392 struct ccp_dma_chan *ccp_dma_chan;
393 struct kmem_cache *dma_cmd_cache;
394 struct kmem_cache *dma_desc_cache;
395
396 /* A counter used to generate job-ids for cmds submitted to the CCP
397 */
398 atomic_t current_id ____cacheline_aligned;
399
400 /* The v3 CCP uses key storage blocks (SB) to maintain context for
401 * certain operations. To prevent multiple cmds from using the same
402 * SB range a command queue reserves an SB range for the duration of
403 * the cmd. Each queue, will however, reserve 2 SB blocks for
404 * operations that only require single SB entries (eg. AES context/iv
405 * and key) in order to avoid allocation contention. This will reserve
406 * at most 10 SB entries, leaving 40 SB entries available for dynamic
407 * allocation.
408 *
409 * The v5 CCP Local Storage Block (LSB) is broken up into 8
410 * memrory ranges, each of which can be enabled for access by one
411 * or more queues. Device initialization takes this into account,
412 * and attempts to assign one region for exclusive use by each
413 * available queue; the rest are then aggregated as "public" use.
414 * If there are fewer regions than queues, all regions are shared
415 * amongst all queues.
416 */
417 struct mutex sb_mutex ____cacheline_aligned;
418 DECLARE_BITMAP(sb, KSB_COUNT);
419 wait_queue_head_t sb_queue;
420 unsigned int sb_avail;
421 unsigned int sb_count;
422 u32 sb_start;
423
424 /* Bitmap of shared LSBs, if any */
425 DECLARE_BITMAP(lsbmap, SLSB_MAP_SIZE);
426
427 /* Suspend support */
428 unsigned int suspending;
429 wait_queue_head_t suspend_queue;
430
431 /* DMA caching attribute support */
432 unsigned int axcache;
433
434 /* Device Statistics */
435 unsigned long total_interrupts;
436
437 /* DebugFS info */
438 struct dentry *debugfs_instance;
439 };
440
441 enum ccp_memtype {
442 CCP_MEMTYPE_SYSTEM = 0,
443 CCP_MEMTYPE_SB,
444 CCP_MEMTYPE_LOCAL,
445 CCP_MEMTYPE__LAST,
446 };
447 #define CCP_MEMTYPE_LSB CCP_MEMTYPE_KSB
448
449
450 struct ccp_dma_info {
451 dma_addr_t address;
452 unsigned int offset;
453 unsigned int length;
454 enum dma_data_direction dir;
455 } __packed __aligned(4);
456
457 struct ccp_dm_workarea {
458 struct device *dev;
459 struct dma_pool *dma_pool;
460
461 u8 *address;
462 struct ccp_dma_info dma;
463 unsigned int length;
464 };
465
466 struct ccp_sg_workarea {
467 struct scatterlist *sg;
468 int nents;
469 unsigned int sg_used;
470
471 struct scatterlist *dma_sg;
472 struct scatterlist *dma_sg_head;
473 struct device *dma_dev;
474 unsigned int dma_count;
475 enum dma_data_direction dma_dir;
476
477 u64 bytes_left;
478 };
479
480 struct ccp_data {
481 struct ccp_sg_workarea sg_wa;
482 struct ccp_dm_workarea dm_wa;
483 };
484
485 struct ccp_mem {
486 enum ccp_memtype type;
487 union {
488 struct ccp_dma_info dma;
489 u32 sb;
490 } u;
491 };
492
493 struct ccp_aes_op {
494 enum ccp_aes_type type;
495 enum ccp_aes_mode mode;
496 enum ccp_aes_action action;
497 unsigned int size;
498 };
499
500 struct ccp_xts_aes_op {
501 enum ccp_aes_type type;
502 enum ccp_aes_action action;
503 enum ccp_xts_aes_unit_size unit_size;
504 };
505
506 struct ccp_des3_op {
507 enum ccp_des3_type type;
508 enum ccp_des3_mode mode;
509 enum ccp_des3_action action;
510 };
511
512 struct ccp_sha_op {
513 enum ccp_sha_type type;
514 u64 msg_bits;
515 };
516
517 struct ccp_rsa_op {
518 u32 mod_size;
519 u32 input_len;
520 };
521
522 struct ccp_passthru_op {
523 enum ccp_passthru_bitwise bit_mod;
524 enum ccp_passthru_byteswap byte_swap;
525 };
526
527 struct ccp_ecc_op {
528 enum ccp_ecc_function function;
529 };
530
531 struct ccp_op {
532 struct ccp_cmd_queue *cmd_q;
533
534 u32 jobid;
535 u32 ioc;
536 u32 soc;
537 u32 sb_key;
538 u32 sb_ctx;
539 u32 init;
540 u32 eom;
541
542 struct ccp_mem src;
543 struct ccp_mem dst;
544 struct ccp_mem exp;
545
546 union {
547 struct ccp_aes_op aes;
548 struct ccp_xts_aes_op xts;
549 struct ccp_des3_op des3;
550 struct ccp_sha_op sha;
551 struct ccp_rsa_op rsa;
552 struct ccp_passthru_op passthru;
553 struct ccp_ecc_op ecc;
554 } u;
555 };
556
ccp_addr_lo(struct ccp_dma_info * info)557 static inline u32 ccp_addr_lo(struct ccp_dma_info *info)
558 {
559 return lower_32_bits(info->address + info->offset);
560 }
561
ccp_addr_hi(struct ccp_dma_info * info)562 static inline u32 ccp_addr_hi(struct ccp_dma_info *info)
563 {
564 return upper_32_bits(info->address + info->offset) & 0x0000ffff;
565 }
566
567 /**
568 * descriptor for version 5 CPP commands
569 * 8 32-bit words:
570 * word 0: function; engine; control bits
571 * word 1: length of source data
572 * word 2: low 32 bits of source pointer
573 * word 3: upper 16 bits of source pointer; source memory type
574 * word 4: low 32 bits of destination pointer
575 * word 5: upper 16 bits of destination pointer; destination memory type
576 * word 6: low 32 bits of key pointer
577 * word 7: upper 16 bits of key pointer; key memory type
578 */
579 struct dword0 {
580 unsigned int soc:1;
581 unsigned int ioc:1;
582 unsigned int rsvd1:1;
583 unsigned int init:1;
584 unsigned int eom:1; /* AES/SHA only */
585 unsigned int function:15;
586 unsigned int engine:4;
587 unsigned int prot:1;
588 unsigned int rsvd2:7;
589 };
590
591 struct dword3 {
592 unsigned int src_hi:16;
593 unsigned int src_mem:2;
594 unsigned int lsb_cxt_id:8;
595 unsigned int rsvd1:5;
596 unsigned int fixed:1;
597 };
598
599 union dword4 {
600 u32 dst_lo; /* NON-SHA */
601 u32 sha_len_lo; /* SHA */
602 };
603
604 union dword5 {
605 struct {
606 unsigned int dst_hi:16;
607 unsigned int dst_mem:2;
608 unsigned int rsvd1:13;
609 unsigned int fixed:1;
610 } fields;
611 u32 sha_len_hi;
612 };
613
614 struct dword7 {
615 unsigned int key_hi:16;
616 unsigned int key_mem:2;
617 unsigned int rsvd1:14;
618 };
619
620 struct ccp5_desc {
621 struct dword0 dw0;
622 u32 length;
623 u32 src_lo;
624 struct dword3 dw3;
625 union dword4 dw4;
626 union dword5 dw5;
627 u32 key_lo;
628 struct dword7 dw7;
629 };
630
631 void ccp_add_device(struct ccp_device *ccp);
632 void ccp_del_device(struct ccp_device *ccp);
633
634 extern void ccp_log_error(struct ccp_device *, unsigned int);
635
636 struct ccp_device *ccp_alloc_struct(struct sp_device *sp);
637 bool ccp_queues_suspended(struct ccp_device *ccp);
638 int ccp_cmd_queue_thread(void *data);
639 int ccp_trng_read(struct hwrng *rng, void *data, size_t max, bool wait);
640
641 int ccp_run_cmd(struct ccp_cmd_queue *cmd_q, struct ccp_cmd *cmd);
642
643 int ccp_register_rng(struct ccp_device *ccp);
644 void ccp_unregister_rng(struct ccp_device *ccp);
645 int ccp_dmaengine_register(struct ccp_device *ccp);
646 void ccp_dmaengine_unregister(struct ccp_device *ccp);
647
648 void ccp5_debugfs_setup(struct ccp_device *ccp);
649 void ccp5_debugfs_destroy(void);
650
651 /* Structure for computation functions that are device-specific */
652 struct ccp_actions {
653 int (*aes)(struct ccp_op *);
654 int (*xts_aes)(struct ccp_op *);
655 int (*des3)(struct ccp_op *);
656 int (*sha)(struct ccp_op *);
657 int (*rsa)(struct ccp_op *);
658 int (*passthru)(struct ccp_op *);
659 int (*ecc)(struct ccp_op *);
660 u32 (*sballoc)(struct ccp_cmd_queue *, unsigned int);
661 void (*sbfree)(struct ccp_cmd_queue *, unsigned int, unsigned int);
662 unsigned int (*get_free_slots)(struct ccp_cmd_queue *);
663 int (*init)(struct ccp_device *);
664 void (*destroy)(struct ccp_device *);
665 irqreturn_t (*irqhandler)(int, void *);
666 };
667
668 extern const struct ccp_vdata ccpv3_platform;
669 extern const struct ccp_vdata ccpv3;
670 extern const struct ccp_vdata ccpv5a;
671 extern const struct ccp_vdata ccpv5b;
672
673 #endif
674