- Acc - mailweb.openeuler.org

UADK周例会
by openEuler conference 19 Jul '22

19 Jul '22

您好！ sig-AccLib SIG 邀请您参加 2022-07-20 11:00 召开的Zoom会议会议主题：UADK周例会会议内容： zoom会议，ID: 91879279131 固定的会议链接：https://linaro-org.zoom.us/j/91879279131 1.进展讨论 2.自由讨论会议链接：https://us06web.zoom.us/j/89055215704?pwd=NkR6SE1HVGpXMHI5MEZoUGJvbUlsQT09 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 温馨提醒：建议接入会议后修改参会人的姓名，也可以使用您在gitee.com的ID 更多资讯尽在：https://openeuler.org/zh/ Hello! openEuler sig-AccLib SIG invites you to attend the Zoom conference will be held at 2022-07-20 11:00, The subject of the conference is UADK周例会, Summary: zoom会议，ID: 91879279131 固定的会议链接：https://linaro-org.zoom.us/j/91879279131 1.进展讨论 2.自由讨论 You can join the meeting at https://us06web.zoom.us/j/89055215704?pwd=NkR6SE1HVGpXMHI5MEZoUGJvbUlsQT09. Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings. Note: You are advised to change the participant name after joining the conference or use your ID at gitee.com. More information: https://openeuler.org/en/

1 0

[PATCH 1/6] uadk_tool: Update some new test features
by Longfang Liu 18 Jul '22

18 Jul '22

1. Modify part of the repeated code macro. 2. Add new random number generation method. 3. Add new test parameters, support test algorithm query and address pre-step function. Signed-off-by: Longfang Liu <liulongfang(a)huawei.com> --- uadk_tool/benchmark/sec_soft_benchmark.c | 7 +- uadk_tool/benchmark/sec_soft_benchmark.h | 1 - uadk_tool/benchmark/sec_uadk_benchmark.c | 65 ++++---- uadk_tool/benchmark/sec_wd_benchmark.c | 192 +++++++---------------- uadk_tool/benchmark/uadk_benchmark.c | 115 ++++++++------ uadk_tool/benchmark/uadk_benchmark.h | 20 ++- 6 files changed, 167 insertions(+), 233 deletions(-) diff --git a/uadk_tool/benchmark/sec_soft_benchmark.c b/uadk_tool/benchmark/sec_soft_benchmark.c index 2f6bccb..a622329 100644 --- a/uadk_tool/benchmark/sec_soft_benchmark.c +++ b/uadk_tool/benchmark/sec_soft_benchmark.c @@ -56,11 +56,8 @@ typedef struct soft_jobs_res { u32 use_engine; } jobs_data; -#define MAX_POOL_LENTH 4096 #define MAX_IVK_LENTH 64 #define DEF_IVK_DATA 0xAA -#define MAX_TRY_CNT 5000 -#define SEND_USLEEP 100 static unsigned int g_thread_num; static unsigned int g_ctxnum; @@ -613,7 +610,7 @@ static void *sec_soft_async_run(void *arg) exit_pause: ASYNC_WAIT_CTX_free(waitctx); - add_recv_data(count); + add_recv_data(count, g_pktlen); return NULL; } @@ -802,7 +799,7 @@ static void *sec_soft_sync_run(void *arg) break; } - add_recv_data(count); + add_recv_data(count, g_pktlen); return NULL; } diff --git a/uadk_tool/benchmark/sec_soft_benchmark.h b/uadk_tool/benchmark/sec_soft_benchmark.h index 8331eb5..c1374d6 100644 --- a/uadk_tool/benchmark/sec_soft_benchmark.h +++ b/uadk_tool/benchmark/sec_soft_benchmark.h @@ -4,6 +4,5 @@ #include "uadk_benchmark.h" - extern int sec_soft_benchmark(struct acc_option *options); #endif /* SEC_SOFT_BENCHMARK_H */ diff --git a/uadk_tool/benchmark/sec_uadk_benchmark.c b/uadk_tool/benchmark/sec_uadk_benchmark.c index 6eeee12..09cd8f4 100644 --- a/uadk_tool/benchmark/sec_uadk_benchmark.c +++ b/uadk_tool/benchmark/sec_uadk_benchmark.c @@ -10,6 +10,8 @@ #include "include/wd_sched.h" #define SEC_TST_PRT printf +#define MAX_IVK_LENTH 64 +#define DEF_IVK_DATA 0xAA struct uadk_bd { u8 *src; @@ -37,18 +39,12 @@ typedef struct uadk_thread_res { u32 td_id; } thread_data; -#define MAX_POOL_LENTH 4096 -#define MAX_IVK_LENTH 64 -#define DEF_IVK_DATA 0xAA -#define MAX_TRY_CNT 5000 -#define SEND_USLEEP 100 - static struct wd_ctx_config g_ctx_cfg; static struct wd_sched *g_sched; static unsigned int g_thread_num; static unsigned int g_ctxnum; +static unsigned int g_prefetch; static unsigned int g_pktlen; -static struct sched_params g_param; static void *cipher_async_cb(struct wd_cipher_req *req, void *data) { @@ -349,6 +345,7 @@ static int sec_uadk_param_parse(thread_data *tddata, struct acc_option *options) static int init_ctx_config(char *alg, int subtype, int mode) { struct uacce_dev_list *list; + struct sched_params param; int i, max_node; int ret = 0; @@ -385,7 +382,7 @@ static int init_ctx_config(char *alg, int subtype, int mode) break; default: SEC_TST_PRT("Fail to parse alg subtype!\n"); - return -EINVAL; + goto out; } if (!g_sched) { SEC_TST_PRT("Fail to alloc sched!\n"); @@ -397,12 +394,12 @@ static int init_ctx_config(char *alg, int subtype, int mode) list->dev->numa_id = 0; g_sched->name = SCHED_SINGLE; - g_param.numa_id = list->dev->numa_id; - g_param.type = 0; - g_param.mode = mode; - g_param.begin = 0; - g_param.end = g_ctxnum - 1; - ret = wd_sched_rr_instance(g_sched, &g_param); + param.numa_id = list->dev->numa_id; + param.type = 0; + param.mode = mode; + param.begin = 0; + param.end = g_ctxnum - 1; + ret = wd_sched_rr_instance(g_sched, &param); if (ret) { SEC_TST_PRT("Fail to fill sched data!\n"); goto out; @@ -419,6 +416,8 @@ static int init_ctx_config(char *alg, int subtype, int mode) case DIGEST_TYPE: ret = wd_digest_init(&g_ctx_cfg, g_sched); break; + default: + goto out; } if (ret) { SEC_TST_PRT("Fail to cipher ctx!\n"); @@ -461,18 +460,13 @@ static void uninit_ctx_config(int subtype) wd_sched_rr_release(g_sched); } -int init_uadk_bd_pool(void) +static int init_uadk_bd_pool(void) { unsigned long step; - int fill_size; int i, j; // make the block not align to 4K step = sizeof(char) * g_pktlen * 2; - if (g_pktlen > MAX_IVK_LENTH) - fill_size = MAX_IVK_LENTH; - else - fill_size = g_pktlen; g_uadk_pool.iv = malloc(g_thread_num * MAX_IVK_LENTH * sizeof(char)); g_uadk_pool.key = malloc(g_thread_num * MAX_IVK_LENTH * sizeof(char)); @@ -498,7 +492,9 @@ int init_uadk_bd_pool(void) if (!g_uadk_pool.pool[i].bds[j].dst) goto malloc_error3; - get_rand_data(g_uadk_pool.pool[i].bds[j].src, fill_size); + get_rand_data(g_uadk_pool.pool[i].bds[j].src, g_pktlen); + if (g_prefetch) + get_rand_data(g_uadk_pool.pool[i].bds[j].dst, g_pktlen); } } } @@ -526,12 +522,13 @@ malloc_error1: free(g_uadk_pool.iv); free(g_uadk_pool.key); + free(g_uadk_pool.mac); SEC_TST_PRT("init uadk bd pool alloc failed!\n"); return -ENOMEM; } -void free_uadk_bd_pool(void) +static void free_uadk_bd_pool(void) { int i, j; @@ -550,21 +547,22 @@ void free_uadk_bd_pool(void) free(g_uadk_pool.iv); free(g_uadk_pool.key); + free(g_uadk_pool.mac); } /*-------------------------------uadk benchmark main code-------------------------------------*/ -void *sec_uadk_poll(void *data) +static void *sec_uadk_poll(void *data) { typedef int (*poll_ctx)(__u32 idx, __u32 expt, __u32 *count); poll_ctx uadk_poll_ctx = NULL; thread_data *pdata = (thread_data *)data; u32 expt = ACC_QUEUE_SIZE * g_thread_num; u32 id = pdata->td_id; - u32 last_time = 2; /* poll need one more recv time */ + u32 last_time = 2; // poll need one more recv time u32 count = 0; u32 recv = 0; - int ret; + int ret; if (id > g_ctxnum) return NULL; @@ -586,7 +584,6 @@ void *sec_uadk_poll(void *data) while (last_time) { ret = uadk_poll_ctx(id, expt, &recv); - // SEC_TST_PRT("expt %u, poll %d recv: %u!\n", expt, i, recv); count += recv; recv = 0; if (unlikely(ret != -WD_EAGAIN && ret < 0)) { @@ -599,7 +596,7 @@ void *sec_uadk_poll(void *data) } recv_error: - add_recv_data(count); + add_recv_data(count, g_pktlen); return NULL; } @@ -636,7 +633,6 @@ static void *sec_uadk_async_run(void *arg) case CIPHER_TYPE: cipher_setup.alg = pdata->alg; cipher_setup.mode = pdata->mode; - cipher_setup.sched_param = (void *)&g_param; h_sess = wd_cipher_alloc_sess(&cipher_setup); if (!h_sess) return NULL; @@ -682,7 +678,6 @@ static void *sec_uadk_async_run(void *arg) case AEAD_TYPE: // just ccm and gcm aead_setup.calg = pdata->alg; aead_setup.cmode = pdata->mode; - aead_setup.sched_param = (void *)&g_param; h_sess = wd_aead_alloc_sess(&aead_setup); if (!h_sess) return NULL; @@ -740,7 +735,6 @@ static void *sec_uadk_async_run(void *arg) case DIGEST_TYPE: digest_setup.alg = pdata->alg; digest_setup.mode = pdata->mode; // digest mode is optype - digest_setup.sched_param = (void *)&g_param; h_sess = wd_digest_alloc_sess(&digest_setup); if (!h_sess) return NULL; @@ -820,7 +814,6 @@ static void *sec_uadk_sync_run(void *arg) case CIPHER_TYPE: cipher_setup.alg = pdata->alg; cipher_setup.mode = pdata->mode; - cipher_setup.sched_param = (void *)&g_param; h_sess = wd_cipher_alloc_sess(&cipher_setup); if (!h_sess) return NULL; @@ -845,7 +838,7 @@ static void *sec_uadk_sync_run(void *arg) creq.src = uadk_pool->bds[i].src; creq.dst = uadk_pool->bds[i].dst; ret = wd_do_cipher_sync(h_sess, &creq); - if (ret || creq.state) + if ((ret < 0 && ret != -WD_EBUSY) || creq.state) break; count++; if (get_run_state() == 0) @@ -856,7 +849,6 @@ static void *sec_uadk_sync_run(void *arg) case AEAD_TYPE: // just ccm and gcm aead_setup.calg = pdata->alg; aead_setup.cmode = pdata->mode; - aead_setup.sched_param = (void *)&g_param; h_sess = wd_aead_alloc_sess(&aead_setup); if (!h_sess) return NULL; @@ -905,7 +897,6 @@ static void *sec_uadk_sync_run(void *arg) case DIGEST_TYPE: digest_setup.alg = pdata->alg; digest_setup.mode = pdata->mode; // digest mode is optype - digest_setup.sched_param = (void *)&g_param; h_sess = wd_digest_alloc_sess(&digest_setup); if (!h_sess) return NULL; @@ -939,7 +930,7 @@ static void *sec_uadk_sync_run(void *arg) break; } - add_recv_data(count); + add_recv_data(count, g_pktlen); return NULL; } @@ -982,7 +973,6 @@ int sec_uadk_sync_threads(struct acc_option *options) sync_error: return ret; - } int sec_uadk_async_threads(struct acc_option *options) @@ -998,10 +988,10 @@ int sec_uadk_async_threads(struct acc_option *options) if (ret) return ret; - /* poll thread */ for (i = 0; i < g_ctxnum; i++) { threads_args[i].subtype = threads_option.subtype; threads_args[i].td_id = i; + /* poll thread */ ret = pthread_create(&pollid[i], NULL, sec_uadk_poll, &threads_args[i]); if (ret) { SEC_TST_PRT("Create poll thread fail!\n"); @@ -1053,6 +1043,7 @@ int sec_uadk_benchmark(struct acc_option *options) g_thread_num = options->threads; g_pktlen = options->pktlen; g_ctxnum = options->ctxnums; + g_prefetch = options->prefetch; if (options->optype > WD_CIPHER_DECRYPTION) { SEC_TST_PRT("SEC optype error: %u\n", options->optype); return -EINVAL; diff --git a/uadk_tool/benchmark/sec_wd_benchmark.c b/uadk_tool/benchmark/sec_wd_benchmark.c index 2e9c55d..d7cfe75 100644 --- a/uadk_tool/benchmark/sec_wd_benchmark.c +++ b/uadk_tool/benchmark/sec_wd_benchmark.c @@ -11,6 +11,9 @@ #include "v1/wd_util.h" #define SEC_TST_PRT printf +#define MAX_IVK_LENTH 64 +#define DEF_IVK_DATA 0xAA +#define SQE_SIZE 128 typedef struct wd_thread_res { u32 subtype; @@ -32,57 +35,31 @@ struct thread_bd_res { struct thread_queue_res { struct thread_bd_res *bd_res; -} g_thread_queue; +}; struct wcrypto_async_tag { void *ctx; - char *out_buf; int thread_id; int cnt; }; -#define MAX_IVK_LENTH 64 -#define DEF_IVK_DATA 0xAA -#define MAX_TRY_CNT 5000 -#define SEND_USLEEP 100 -#define TEST_MAX_THRD 128 -#define SQE_SIZE 128 -#define MAX_BLOCK_NM 4096 - +static struct thread_queue_res g_thread_queue; static unsigned int g_thread_num; -static unsigned int g_ctxnum; static unsigned int g_pktlen; static void *cipher_async_cb(void *message, void *cipher_tag) { - struct wcrypto_async_tag *async_tag = (struct wcrypto_async_tag *)cipher_tag; - struct wcrypto_cipher_msg *req = (struct wcrypto_cipher_msg *)message; - - // no-sva data copy from uadk to user - memcpy(async_tag->out_buf, req->out, g_pktlen); - return NULL; } -static void *aead_async_cb(void *message, void *aead_tag) +static void *aead_async_cb(void *message, void *cipher_tag) { - struct wcrypto_async_tag *async_tag = (struct wcrypto_async_tag *)aead_tag; - struct wcrypto_aead_msg *req = (struct wcrypto_aead_msg *)message; - - // no-sva data copy from uadk to user - memcpy(async_tag->out_buf, req->out, g_pktlen); - return NULL; } static void *digest_async_cb(void *message, void *digest_tag) { - struct wcrypto_async_tag *async_tag = (struct wcrypto_async_tag *)digest_tag; - struct wcrypto_digest_msg *req = (struct wcrypto_digest_msg *)message; - - // no-sva data copy from uadk to user - memcpy(async_tag->out_buf, req->out, 16); - + // struct WCRYPTO_req *req = (struct WCRYPTO_req *)data; return NULL; } @@ -396,9 +373,9 @@ static int init_wd_queue(struct acc_option *options) // use no-sva pbuffer memset(&blksetup, 0, sizeof(blksetup)); blksetup.block_size = g_pktlen + SQE_SIZE; //aead need mac and aad out - blksetup.block_num = MAX_BLOCK_NM * 4; //set pool inv + key + in + out + blksetup.block_num = MAX_BLOCK_NM; //set pool inv + key + in + out blksetup.align_size = SQE_SIZE; - SEC_TST_PRT("create pool memory: %d KB\n", (MAX_BLOCK_NM * blksetup.block_size) >> 10); + // SEC_TST_PRT("create pool memory: %d KB\n", (MAX_BLOCK_NM * blksetup.block_size) >> 10); for (j = 0; j < g_thread_num; j++) { g_thread_queue.bd_res[j].pool = wd_blkpool_create(g_thread_queue.bd_res[j].queue, &blksetup); @@ -411,8 +388,8 @@ static int init_wd_queue(struct acc_option *options) // alloc in pbuffer res for (m = 0; m < g_thread_num; m++) { - g_thread_queue.bd_res[m].in = malloc(MAX_BLOCK_NM * sizeof(void *)); - for (idx = 0; idx < MAX_BLOCK_NM; idx++) { + g_thread_queue.bd_res[m].in = malloc(MAX_POOL_LENTH * sizeof(void *)); + for (idx = 0; idx < MAX_POOL_LENTH; idx++) { g_thread_queue.bd_res[m].in[idx] = wd_alloc_blk(g_thread_queue.bd_res[m].pool); if (!g_thread_queue.bd_res[m].in[idx]) { SEC_TST_PRT("create pool %dth in memory fail!\n", m); @@ -427,8 +404,8 @@ static int init_wd_queue(struct acc_option *options) // alloc out pbuffer res for (n = 0; n < g_thread_num; n++) { - g_thread_queue.bd_res[n].out = malloc(MAX_BLOCK_NM * sizeof(void *)); - for (idx = 0; idx < MAX_BLOCK_NM; idx++) { + g_thread_queue.bd_res[n].out = malloc(MAX_POOL_LENTH * sizeof(void *)); + for (idx = 0; idx < MAX_POOL_LENTH; idx++) { g_thread_queue.bd_res[n].out[idx] = wd_alloc_blk(g_thread_queue.bd_res[n].pool); if (!g_thread_queue.bd_res[n].out[idx]) { SEC_TST_PRT("create pool %dth out memory fail!\n", n); @@ -443,8 +420,8 @@ static int init_wd_queue(struct acc_option *options) // alloc iv pbuffer res for (k = 0; k < g_thread_num; k++) { - g_thread_queue.bd_res[k].iv = malloc(MAX_BLOCK_NM * sizeof(void *)); - for (idx = 0; idx < MAX_BLOCK_NM; idx++) { + g_thread_queue.bd_res[k].iv = malloc(MAX_POOL_LENTH * sizeof(void *)); + for (idx = 0; idx < MAX_POOL_LENTH; idx++) { g_thread_queue.bd_res[k].iv[idx] = wd_alloc_blk(g_thread_queue.bd_res[k].pool); if (!g_thread_queue.bd_res[k].iv[idx]) { SEC_TST_PRT("create pool %dth iv memory fail!\n", k); @@ -462,21 +439,21 @@ static int init_wd_queue(struct acc_option *options) iv_err: for (k--; k >= 0; k--) { - for (idx = 0; idx < MAX_BLOCK_NM; idx++) + for (idx = 0; idx < MAX_POOL_LENTH; idx++) wd_free_blk(g_thread_queue.bd_res[k].pool, g_thread_queue.bd_res[k].iv[idx]); free(g_thread_queue.bd_res[k].iv); } out_err: for (n--; n >= 0; n--) { - for (idx = 0; idx < MAX_BLOCK_NM; idx++) + for (idx = 0; idx < MAX_POOL_LENTH; idx++) wd_free_blk(g_thread_queue.bd_res[n].pool, g_thread_queue.bd_res[n].out[idx]); free(g_thread_queue.bd_res[n].out); } in_err: for (m--; m >= 0; m--) { - for (idx = 0; idx < MAX_BLOCK_NM; idx++) + for (idx = 0; idx < MAX_POOL_LENTH; idx++) wd_free_blk(g_thread_queue.bd_res[m].pool, g_thread_queue.bd_res[m].in[idx]); free(g_thread_queue.bd_res[m].in); @@ -498,7 +475,7 @@ static void uninit_wd_queue(void) int i, j, idx; for (i = 0; i < g_thread_num; i++) { - for (idx = 0; idx < MAX_BLOCK_NM; idx++) { + for (idx = 0; idx < MAX_POOL_LENTH; idx++) { wd_free_blk(g_thread_queue.bd_res[i].pool, g_thread_queue.bd_res[i].iv[idx]); wd_free_blk(g_thread_queue.bd_res[i].pool, g_thread_queue.bd_res[i].in[idx]); wd_free_blk(g_thread_queue.bd_res[i].pool, g_thread_queue.bd_res[i].out[idx]); @@ -522,22 +499,22 @@ void *sec_wd_poll(void *data) { typedef int (*poll_ctx)(struct wd_queue *q, unsigned int num); thread_data *pdata = (thread_data *)data; - poll_ctx uadk_poll_ctx = NULL; + poll_ctx wd_poll_ctx = NULL; u32 expt = ACC_QUEUE_SIZE * g_thread_num; - u32 last_time = 2; /* poll need one more recv time */ + u32 last_time = 2; // poll need one more recv time u32 id = pdata->td_id; u32 count = 0; - u32 recv = 0; + int recv = 0; switch(pdata->subtype) { case CIPHER_TYPE: - uadk_poll_ctx = wcrypto_cipher_poll; + wd_poll_ctx = wcrypto_cipher_poll; break; case AEAD_TYPE: - uadk_poll_ctx = wcrypto_aead_poll; + wd_poll_ctx = wcrypto_aead_poll; break; case DIGEST_TYPE: - uadk_poll_ctx = wcrypto_digest_poll; + wd_poll_ctx = wcrypto_digest_poll; break; default: SEC_TST_PRT("<<<<<<async poll interface is NULL!\n"); @@ -548,7 +525,7 @@ void *sec_wd_poll(void *data) return NULL; while (last_time) { - recv = uadk_poll_ctx(g_thread_queue.bd_res[id].queue, expt); + recv = wd_poll_ctx(g_thread_queue.bd_res[id].queue, expt); /* * warpdrive async mode poll easy to 100% with small package. * SEC_TST_PRT("warpdrive poll %d recv: %u!\n", i, recv); @@ -565,7 +542,7 @@ void *sec_wd_poll(void *data) } recv_error: - add_recv_data(count); + add_recv_data(count, g_pktlen); return NULL; } @@ -582,8 +559,6 @@ static void *sec_wd_async_run(void *arg) struct wcrypto_async_tag *tag = NULL; char priv_key[MAX_IVK_LENTH]; struct thread_bd_res *bd_res; - char *src_data_buf = NULL; - char *out_data_buf = NULL; struct wd_queue *queue; void *ctx = NULL; void **res_in; @@ -605,27 +580,13 @@ static void *sec_wd_async_run(void *arg) res_out = bd_res->out; res_iv = bd_res->iv; - /* create user data buffer */ - src_data_buf = malloc(g_pktlen * sizeof(char)); - if (!src_data_buf) - return NULL; - - get_rand_data((u8 *)src_data_buf, g_pktlen); - out_data_buf = malloc(g_pktlen * sizeof(char)); - if (!out_data_buf) { - free(src_data_buf); - return NULL; - } - memset(priv_key, DEF_IVK_DATA, MAX_IVK_LENTH); tag = malloc(sizeof(struct wcrypto_async_tag)); // set the user tag if (!tag) { SEC_TST_PRT("wcrypto async alloc tag fail!\n"); - free(src_data_buf); - free(out_data_buf); return NULL; } - tag->out_buf = out_data_buf; + tag->thread_id = pdata->td_id; switch(pdata->subtype) { case CIPHER_TYPE: @@ -642,7 +603,7 @@ static void *sec_wd_async_run(void *arg) ctx = wcrypto_create_cipher_ctx(queue, &cipher_setup); if (!ctx) { SEC_TST_PRT("wd create cipher ctx fail!\n"); - goto async_err; + return NULL; } tag->ctx = ctx; @@ -650,7 +611,7 @@ static void *sec_wd_async_run(void *arg) if (ret) { SEC_TST_PRT("wd cipher set key fail!\n"); wcrypto_del_cipher_ctx(ctx); - goto async_err; + return NULL; } if (queue->capa.priv.direction == 0) @@ -672,9 +633,6 @@ static void *sec_wd_async_run(void *arg) if (get_run_state() == 0) break; - // no-sva data copy to uadk - memcpy(copdata.in, src_data_buf, g_pktlen); - ret = wcrypto_do_cipher(ctx, &copdata, (void *)tag); if (ret == -WD_EBUSY) { usleep(SEND_USLEEP * try_cnt); @@ -687,7 +645,7 @@ static void *sec_wd_async_run(void *arg) } count++; - i = count % MAX_BLOCK_NM; + i = count % MAX_POOL_LENTH; tag->cnt = i; try_cnt = 0; copdata.in = res_in[i]; @@ -710,7 +668,7 @@ static void *sec_wd_async_run(void *arg) ctx = wcrypto_create_aead_ctx(queue, &aead_setup); if (!ctx) { SEC_TST_PRT("wd create aead ctx fail!\n"); - goto async_err; + return NULL; } tag->ctx = ctx; @@ -718,7 +676,7 @@ static void *sec_wd_async_run(void *arg) if (ret) { SEC_TST_PRT("wd aead set key fail!\n"); wcrypto_del_aead_ctx(ctx); - goto async_err; + return NULL; } authsize = 16; //set defaut size @@ -726,7 +684,7 @@ static void *sec_wd_async_run(void *arg) if (ret) { SEC_TST_PRT("set authsize fail!\n"); wcrypto_del_aead_ctx(ctx); - goto async_err; + return NULL; } if (queue->capa.priv.direction == 0) { @@ -753,9 +711,6 @@ static void *sec_wd_async_run(void *arg) if (get_run_state() == 0) break; - // no-sva data copy to uadk - memcpy(aopdata.in, src_data_buf, g_pktlen); - ret = wcrypto_do_aead(ctx, &aopdata, (void *)tag); if (ret == -WD_EBUSY) { usleep(SEND_USLEEP * try_cnt); @@ -768,7 +723,7 @@ static void *sec_wd_async_run(void *arg) } count++; - i = count % MAX_BLOCK_NM; + i = count % MAX_POOL_LENTH; tag->cnt = i; try_cnt = 0; aopdata.in = res_in[i]; @@ -791,7 +746,7 @@ static void *sec_wd_async_run(void *arg) ctx = wcrypto_create_digest_ctx(queue, &digest_setup); if (!ctx) { SEC_TST_PRT("wd create digest ctx fail!\n"); - goto async_err; + return NULL; } tag->ctx = ctx; @@ -801,7 +756,7 @@ static void *sec_wd_async_run(void *arg) if (ret) { SEC_TST_PRT("wd digest set key fail!\n"); wcrypto_del_digest_ctx(ctx); - goto async_err; + return NULL; } } @@ -818,9 +773,6 @@ static void *sec_wd_async_run(void *arg) if (get_run_state() == 0) break; - // no-sva data copy to uadk - memcpy(dopdata.in, src_data_buf, g_pktlen); - ret = wcrypto_do_digest(ctx, &dopdata, (void *)tag); if (ret == -WD_EBUSY) { usleep(SEND_USLEEP * try_cnt); @@ -833,7 +785,7 @@ static void *sec_wd_async_run(void *arg) } count++; - i = count % MAX_BLOCK_NM; + i = count % MAX_POOL_LENTH; tag->cnt = i; try_cnt = 0; dopdata.in = res_in[i]; @@ -846,7 +798,7 @@ static void *sec_wd_async_run(void *arg) add_send_complete(); while (1) { - if (get_recv_time() == g_thread_num) // wait Async mode finish recv + if (get_recv_time() > 0) // wait Async mode finish recv break; usleep(SEND_USLEEP); } @@ -863,10 +815,6 @@ static void *sec_wd_async_run(void *arg) break; } -async_err: - free(tag); - free(src_data_buf); - free(out_data_buf); return NULL; } @@ -882,8 +830,6 @@ static void *sec_wd_sync_run(void *arg) char priv_key[MAX_IVK_LENTH]; struct thread_bd_res *bd_res; struct wd_queue *queue; - char *src_data_buf = NULL; - char *out_data_buf = NULL; void *ctx = NULL; void *tag = NULL; void **res_in; @@ -905,18 +851,6 @@ static void *sec_wd_sync_run(void *arg) res_out = bd_res->out; res_iv = bd_res->iv; - /* create user data buffer */ - src_data_buf = malloc(g_pktlen * sizeof(char)); - if (!src_data_buf) - return NULL; - - get_rand_data((u8 *)src_data_buf, g_pktlen); - out_data_buf = malloc(g_pktlen * sizeof(char)); - if (!out_data_buf) { - free(src_data_buf); - return NULL; - } - memset(priv_key, DEF_IVK_DATA, MAX_IVK_LENTH); switch(pdata->subtype) { @@ -933,14 +867,14 @@ static void *sec_wd_sync_run(void *arg) ctx = wcrypto_create_cipher_ctx(queue, &cipher_setup); if (!ctx) { SEC_TST_PRT("wd create cipher ctx fail!\n"); - goto sync_err; + return NULL; } ret = wcrypto_set_cipher_key(ctx, (__u8*)priv_key, (__u16)pdata->keysize); if (ret) { SEC_TST_PRT("wd cipher set key fail!\n"); wcrypto_del_cipher_ctx(ctx); - goto sync_err; + return NULL; } if (queue->capa.priv.direction == 0) @@ -961,9 +895,6 @@ static void *sec_wd_sync_run(void *arg) if (get_run_state() == 0) break; - // no-sva data copy to uadk - memcpy(copdata.in, src_data_buf, g_pktlen); - ret = wcrypto_do_cipher(ctx, &copdata, tag); if (ret == -WD_EBUSY) { usleep(SEND_USLEEP * try_cnt); @@ -977,13 +908,10 @@ static void *sec_wd_sync_run(void *arg) count++; try_cnt = 0; - i = count % MAX_BLOCK_NM; + i = count % MAX_POOL_LENTH; copdata.in = res_in[i]; copdata.out = res_out[i]; copdata.iv = res_iv[i]; - - // no-sva data copy from uadk to user - memcpy(out_data_buf, copdata.out, g_pktlen); } wcrypto_del_cipher_ctx(ctx); @@ -1001,14 +929,14 @@ static void *sec_wd_sync_run(void *arg) ctx = wcrypto_create_aead_ctx(queue, &aead_setup); if (!ctx) { SEC_TST_PRT("wd create aead ctx fail!\n"); - goto sync_err; + return NULL; } ret = wcrypto_set_aead_ckey(ctx, (__u8*)priv_key, (__u16)pdata->keysize); if (ret) { SEC_TST_PRT("wd aead set key fail!\n"); wcrypto_del_aead_ctx(ctx); - goto sync_err; + return NULL; } authsize = 16; //set defaut size @@ -1016,7 +944,7 @@ static void *sec_wd_sync_run(void *arg) if (ret) { SEC_TST_PRT("set authsize fail!\n"); wcrypto_del_aead_ctx(ctx); - goto sync_err; + return NULL; } if (queue->capa.priv.direction == 0) { @@ -1042,9 +970,6 @@ static void *sec_wd_sync_run(void *arg) if (get_run_state() == 0) break; - // no-sva data copy to uadk - memcpy(aopdata.in, src_data_buf, g_pktlen); - ret = wcrypto_do_aead(ctx, &aopdata, tag); if (ret == -WD_EBUSY) { usleep(SEND_USLEEP * try_cnt); @@ -1058,13 +983,10 @@ static void *sec_wd_sync_run(void *arg) count++; try_cnt = 0; - i = count % MAX_BLOCK_NM; + i = count % MAX_POOL_LENTH; aopdata.in = res_in[i]; aopdata.out = res_out[i]; aopdata.iv = res_iv[i]; - - // no-sva data copy from uadk to user - memcpy(out_data_buf, aopdata.out, g_pktlen); } wcrypto_del_aead_ctx(ctx); @@ -1082,7 +1004,7 @@ static void *sec_wd_sync_run(void *arg) ctx = wcrypto_create_digest_ctx(queue, &digest_setup); if (!ctx) { SEC_TST_PRT("wd create digest ctx fail!\n"); - goto sync_err; + return NULL; } if (digest_setup.mode == WCRYPTO_DIGEST_HMAC) { @@ -1091,7 +1013,7 @@ static void *sec_wd_sync_run(void *arg) if (ret) { SEC_TST_PRT("wd digest set key fail!\n"); wcrypto_del_digest_ctx(ctx); - goto sync_err; + return NULL; } } @@ -1107,9 +1029,6 @@ static void *sec_wd_sync_run(void *arg) if (get_run_state() == 0) break; - // no-sva data copy to uadk - memcpy(dopdata.in, src_data_buf, g_pktlen); - ret = wcrypto_do_digest(ctx, &dopdata, (void *)tag); if (ret == -WD_EBUSY) { usleep(SEND_USLEEP * try_cnt); @@ -1123,22 +1042,16 @@ static void *sec_wd_sync_run(void *arg) count++; try_cnt = 0; - i = count % MAX_BLOCK_NM; + i = count % MAX_POOL_LENTH; dopdata.in = res_in[i]; - dopdata.out = res_out[i]; - - // no-sva data copy from uadk to user - memcpy(out_data_buf, dopdata.out, 16); + dopdata.out = res_out[i]; } wcrypto_del_digest_ctx(ctx); break; } - add_recv_data(count); + add_recv_data(count, g_pktlen); -sync_err: - free(src_data_buf); - free(out_data_buf); return NULL; } @@ -1180,6 +1093,7 @@ int sec_wd_sync_threads(struct acc_option *options) sync_error: return ret; + } int sec_wd_async_threads(struct acc_option *options) @@ -1249,7 +1163,6 @@ int sec_wd_benchmark(struct acc_option *options) g_thread_num = options->threads; g_pktlen = options->pktlen; - g_ctxnum = options->ctxnums; if (options->optype > WCRYPTO_CIPHER_DECRYPTION) { SEC_TST_PRT("SEC optype error: %u\n", options->optype); return -EINVAL; @@ -1273,3 +1186,4 @@ int sec_wd_benchmark(struct acc_option *options) return 0; } + diff --git a/uadk_tool/benchmark/uadk_benchmark.c b/uadk_tool/benchmark/uadk_benchmark.c index 84941d5..2904cba 100644 --- a/uadk_tool/benchmark/uadk_benchmark.c +++ b/uadk_tool/benchmark/uadk_benchmark.c @@ -10,19 +10,19 @@ #define BYTES_TO_KB 10 #define TABLE_SPACE_SIZE 8 -#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) /*----------------------------------------head struct--------------------------------------------------------*/ static unsigned int g_run_state = 1; static pthread_mutex_t acc_mutex = PTHREAD_MUTEX_INITIALIZER; static struct _recv_data { + double pkg_len; u64 send_cnt; u64 recv_cnt; u32 send_times; u32 recv_times; } g_recv_data; -/* SVA mode and NOSVA mode change need re_insmode driver ko */ +/* SVA mode and NOSVA mode change need re_insmod driver ko */ enum test_type { SVA_MODE = 0x1, NOSVA_MODE = 0x2, @@ -63,7 +63,7 @@ static struct acc_alg_item alg_options[] = { {"rsa-2048-crt", RSA_2048_CRT}, {"rsa-3072-crt", RSA_3072_CRT}, {"rsa-4096-crt", RSA_4096_CRT}, - {"dh-768 ", DH_768}, + {"dh-768", DH_768}, {"dh-1024", DH_1024}, {"dh-1536", DH_1536}, {"dh-2048", DH_2048}, @@ -131,10 +131,14 @@ void add_send_complete(void) __atomic_add_fetch(&g_recv_data.send_times, 1, __ATOMIC_RELAXED); } -void add_recv_data(u32 cnt) +void add_recv_data(u32 cnt, u32 pkglen) { pthread_mutex_lock(&acc_mutex); g_recv_data.recv_cnt += cnt; + if (g_recv_data.pkg_len == 0) + g_recv_data.pkg_len = pkglen; + else + g_recv_data.pkg_len = ((double)pkglen + g_recv_data.pkg_len) / 2; g_recv_data.recv_times++; pthread_mutex_unlock(&acc_mutex); } @@ -148,6 +152,7 @@ void init_recv_data(void) { g_recv_data.send_cnt = 0; g_recv_data.recv_cnt = 0; + g_recv_data.pkg_len = 0.0; g_recv_data.send_times = 0; g_recv_data.recv_times = 0; } @@ -263,28 +268,21 @@ void time_start(u32 seconds) alarm(seconds); } -int get_rand_int(int range) -{ - int randnum; - - if (range <= 0) { - ACC_TST_PRT("rand range error!\n"); - return 1; - } - srand((unsigned)time(NULL) * getpid()); - randnum = rand() % range; - - return randnum; -} - -void get_rand_data(u8 *addr, int size) +void get_rand_data(u8 *addr, u32 size) { + unsigned short rand_state[3] = { + (0xae >> 16) & 0xffff, 0xae & 0xffff, 0x330e}; int i; - srand((unsigned)time(NULL) * getpid()); - for (i = 0; i < size; i++) { - addr[i] = rand() % 0xFF; - } +#if 1 + // only 32bit valid, other 32bit is zero + for (i = 0; i < size >> 3; i++) + *((u64 *)addr + i) = nrand48(rand_state); +#else + // full 64bit valid + for (i = 0; i < size >> 2; i++) + *((u32 *)addr + i) = nrand48(rand_state); +#endif } /*-------------------------------------main code------------------------------------------------------*/ @@ -320,7 +318,7 @@ static void parse_alg_param(struct acc_option *option) case X25519_ALG: snprintf(option->algclass, MAX_ALG_NAME, "%s", "x25519"); option->acctype = HPRE_TYPE; - option->subtype = X22519_TYPE; + option->subtype = X25519_TYPE; break; case X448_ALG: snprintf(option->algclass, MAX_ALG_NAME, "%s", "x448"); @@ -366,8 +364,8 @@ void cal_perfermance_data(struct acc_option *option, u32 sttime) double perfermance; double cpu_rate; u32 ttime = 1000; - u32 perfdata; - u32 perfops; + double perfdata; + double perfops; double ops; u32 ptime; int i, len; @@ -397,13 +395,13 @@ void cal_perfermance_data(struct acc_option *option, u32 sttime) palgname[i] = '\0'; ptime = ptime - sttime; - perfdata = (g_recv_data.recv_cnt * option->pktlen) >> BYTES_TO_KB; - perfops = (g_recv_data.recv_cnt) >> BYTES_TO_KB; - perfermance = (double)perfdata / option->times; - ops = (double)perfops / option->times; + perfdata = g_recv_data.pkg_len * g_recv_data.recv_cnt / 1024.0; + perfops = (double)(g_recv_data.recv_cnt) / 1000.0; + perfermance = perfdata / option->times; + ops = perfops / option->times; cpu_rate = (double)ptime / option->times; ACC_TST_PRT("algname: length: perf: iops: CPU_rate:\n" - "%s %uBytes %.1fKB/s %.1fKops %.2f%%\n", + "%s %-2uBytes %.1fKB/s %.1fKops %.2f%%\n", palgname, option->pktlen, perfermance, ops, cpu_rate); } @@ -419,11 +417,11 @@ static int benchmark_run(struct acc_option *option) ret = sec_wd_benchmark(option); } usleep(20000); - #ifdef WITH_OPENSSL_DIR +#ifdef WITH_OPENSSL_DIR if (option->modetype & SOFT_MODE) { ret = sec_soft_benchmark(option); } - #endif +#endif break; case HPRE_TYPE: break; @@ -447,6 +445,7 @@ static void dump_param(struct acc_option *option) ACC_TST_PRT(" [--ctxnum]: %u\n", option->ctxnums); ACC_TST_PRT(" [--algclass]:%s\n", option->algclass); ACC_TST_PRT(" [--acctype]: %u\n", option->acctype); + ACC_TST_PRT(" [--prefetch]:%u\n", option->prefetch); ACC_TST_PRT(" [--engine]: %s\n", option->engine); } @@ -457,6 +456,7 @@ int acc_benchmark_run(struct acc_option *option) int i, ret = 0; int status; + ACC_TST_PRT("start UADK benchmark test.\n"); parse_alg_param(option); dump_param(option); @@ -558,13 +558,27 @@ static void print_help(void) ACC_TST_PRT(" set the number of threads\n"); ACC_TST_PRT(" [--ctxnum]:\n"); ACC_TST_PRT(" the number of QP queues used by the entire test task\n"); + ACC_TST_PRT(" [--prefetch]:\n"); + ACC_TST_PRT(" in SVA mode, Enable prefetch can reduce page faults and improve performance\n"); ACC_TST_PRT(" [--engine]:\n"); ACC_TST_PRT(" set the test openssl engine\n"); + ACC_TST_PRT(" [--alglist]:\n"); + ACC_TST_PRT(" list the all support alg\n"); ACC_TST_PRT(" [--help] = usage\n"); ACC_TST_PRT("Example\n"); ACC_TST_PRT(" ./uadk_tool benchmark --alg aes-128-cbc --mode sva --opt 0 --sync\n"); ACC_TST_PRT(" --pktlen 1024 --seconds 1 --multi 1 --thread 1 --ctxnum 4\n"); - ACC_TST_PRT("UPDATE:2021-7-28\n"); + ACC_TST_PRT("UPDATE:2022-7-18\n"); +} + +static void print_support_alg(void) +{ + int i; + + ACC_TST_PRT("UADK benchmark supported ALG:\n"); + for (i = 0; i < ALG_MAX; i++) { + ACC_TST_PRT("%s\n", alg_options[i].name); + } } int acc_cmd_parse(int argc, char *argv[], struct acc_option *option) @@ -583,8 +597,10 @@ int acc_cmd_parse(int argc, char *argv[], struct acc_option *option) {"thread", required_argument, 0, 9}, {"multi", required_argument, 0, 10}, {"ctxnum", required_argument, 0, 11}, - {"engine", required_argument, 0, 12}, - {"help", no_argument, 0, 13}, + {"prefetch", no_argument, 0, 12}, + {"engine", required_argument, 0, 13}, + {"alglist", no_argument, 0, 14}, + {"help", no_argument, 0, 15}, {0, 0, 0, 0} }; @@ -593,7 +609,6 @@ int acc_cmd_parse(int argc, char *argv[], struct acc_option *option) if (c == -1) break; - // ACC_TST_PRT("index:%d , optarg name:%s\n", c, optarg); switch (c) { case 2: option->algtype = get_alg_type(optarg); @@ -627,26 +642,36 @@ int acc_cmd_parse(int argc, char *argv[], struct acc_option *option) option->ctxnums = strtol(optarg, NULL, 0); break; case 12: - strcpy(option->engine, optarg); + option->prefetch = 1; break; case 13: - print_help(); + strcpy(option->engine, optarg); break; + case 14: + print_support_alg(); + goto to_exit; + case 15: + print_help(); + goto to_exit; default: ACC_TST_PRT("bad input test parameter!\n"); print_help(); - exit(-1); + goto to_exit; } } + return 0; + +to_exit: + return -EINVAL; } int acc_option_convert(struct acc_option *option) { - if (option->algtype >= ALG_MAX) + if (option->algtype >= ALG_MAX) { + ACC_TST_PRT("invalid: input algname is wrong!\n"); goto param_err; - else if (option->algtype < 0) - option->algtype = AES_128_CBC; + } if (option->modetype >= INVALID_MODE) goto param_err; @@ -657,8 +682,8 @@ int acc_option_convert(struct acc_option *option) /* Min test package size is 64Bytes */ if (option->pktlen > MAX_DATA_SIZE) goto param_err; - else if (option->pktlen < 64) - option->pktlen = 64; + else if (option->pktlen < 16) + option->pktlen = 16; if (option->times > MAX_TIME_SECONDS) { ACC_TST_PRT("uadk benchmark max test times to 128 seconds\n"); diff --git a/uadk_tool/benchmark/uadk_benchmark.h b/uadk_tool/benchmark/uadk_benchmark.h index 2cb3dc0..a344fac 100644 --- a/uadk_tool/benchmark/uadk_benchmark.h +++ b/uadk_tool/benchmark/uadk_benchmark.h @@ -15,7 +15,6 @@ #include <signal.h> #include <linux/random.h> #include <sys/syscall.h> -#include <sys/mman.h> #include <sys/time.h> #define ACC_TST_PRT printf @@ -24,21 +23,30 @@ #define MAX_CTX_NUM 64 #define MAX_TIME_SECONDS 128 #define BYTES_TO_MB 20 -#define MAX_OPT_TYPE 5 +#define MAX_OPT_TYPE 6 #define MAX_DATA_SIZE (15 * 1024 * 1024) #define MAX_ALG_NAME 64 #define ACC_QUEUE_SIZE 1024 +#define MAX_BLOCK_NM 16384 /* BLOCK_NUM must 4 times of POOL_LENTH */ +#define MAX_POOL_LENTH 4096 +#define MAX_TRY_CNT 5000 +#define SEND_USLEEP 100 + typedef unsigned char u8; typedef unsigned int u32; typedef unsigned long long u64; #define SCHED_SINGLE "sched_single" +#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) /** * struct acc_option - Define the test acc app option list. * @algclass: 0:cipher 1:digest * @acctype: The sub alg type, reference func get_cipher_resource. * @syncmode: 0:sync mode 1:async mode + * @modetype: sva, no-sva, soft mode + * @optype: enc/dec, comp/decomp + * @prefetch: write allocated memory to prevent page faults */ struct acc_option { char algname[64]; @@ -56,6 +64,7 @@ struct acc_option { u32 subtype; char engine[64]; u32 engine_flag; + u32 prefetch; }; enum acc_type { @@ -74,7 +83,7 @@ enum alg_type { ECDH_TYPE, ECDSA_TYPE, SM2_TYPE, - X22519_TYPE, + X25519_TYPE, X448_TYPE, }; @@ -164,9 +173,8 @@ extern void cal_perfermance_data(struct acc_option *option, u32 sttime); extern void time_start(u32 seconds); extern int get_run_state(void); extern void set_run_state(int state); -extern int get_rand_int(int range); -extern void get_rand_data(u8 *addr, int size); -extern void add_recv_data(u32 cnt); +extern void get_rand_data(u8 *addr, u32 size); +extern void add_recv_data(u32 cnt, u32 pkglen); extern void add_send_complete(void); extern u32 get_recv_time(void); -- 2.33.0

1 5

[PATCH] digest: fix the fault as using the nginx
by Kai Ye 18 Jul '22

18 Jul '22

Prevent double-free after the private ctx copy is used. Signed-off-by: Kai Ye <yekai13(a)huawei.com> --- src/uadk_digest.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/src/uadk_digest.c b/src/uadk_digest.c index 2e61e80..8127373 100644 --- a/src/uadk_digest.c +++ b/src/uadk_digest.c @@ -240,6 +240,10 @@ static void digest_soft_cleanup(struct digest_priv_ctx *md_ctx) { EVP_MD_CTX *ctx = md_ctx->soft_ctx; + /* Prevent double-free after the copy is used */ + if (md_ctx->copy) + return; + if (ctx != NULL) { if (ctx->md_data) { OPENSSL_free(ctx->md_data); @@ -641,7 +645,9 @@ static int do_digest_sync(struct digest_priv_ctx *priv) { int ret; - /* Fix me: not support switch the soft work as input is lower */ + if (priv->req.in_bytes <= priv->switch_threshold && + priv->state == SEC_DIGEST_INIT) + return 0; ret = wd_do_digest_sync(priv->sess, &priv->req); if (ret) { @@ -743,6 +749,7 @@ static int uadk_e_digest_cleanup(EVP_MD_CTX *ctx) struct digest_priv_ctx *priv = (struct digest_priv_ctx *)EVP_MD_CTX_md_data(ctx); + /* Prevent double-free after the copy is used */ if (!priv || priv->copy) return 1; -- 2.33.0

1 0

[PATCH 1/2] uadk: fix spin lock destroy
by Weili Qian 18 Jul '22

18 Jul '22

Fix missing destroyed lock when initialization fails. Signed-off-by: Wenkai Lin <linwenkai6(a)hisilicon.com> Signed-off-by: Weili Qian <qianweili(a)huawei.com> --- drv/hisi_qm_udrv.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff --git a/drv/hisi_qm_udrv.c b/drv/hisi_qm_udrv.c index 9eaf60c..b4e763d 100644 --- a/drv/hisi_qm_udrv.c +++ b/drv/hisi_qm_udrv.c @@ -335,6 +335,14 @@ err_out: return ret; } +static void hisi_qm_clear_info(struct hisi_qp *qp) +{ + struct hisi_qm_queue_info *q_info = &qp->q_info; + + pthread_spin_destroy(&q_info->lock); + hisi_qm_unset_region(qp->h_ctx, q_info); +} + static int get_free_num(struct hisi_qm_queue_info *q_info) { /* The device should reserve one buffer. */ @@ -374,7 +382,7 @@ handle_t hisi_qm_alloc_qp(struct hisi_qm_priv *config, handle_t ctx) qp->h_sgl_pool = hisi_qm_create_sglpool(HISI_SGL_NUM_IN_BD, HISI_SGE_NUM_IN_SGL); if (!qp->h_sgl_pool) - goto out_qp; + goto free_info; ret = wd_ctx_start(qp->h_ctx); if (ret) @@ -390,6 +398,8 @@ handle_t hisi_qm_alloc_qp(struct hisi_qm_priv *config, handle_t ctx) free_pool: hisi_qm_destroy_sglpool(qp->h_sgl_pool); +free_info: + hisi_qm_clear_info(qp); out_qp: free(qp); out: -- 2.30.0

1 1

[PATCH] engine: add receiving timeout count
by Kai Ye 18 Jul '22

18 Jul '22

Increase the count of packet receiving timeout as doing async jobs. Prevents falling into an infinite loop. Signed-off-by: Kai Ye <yekai13(a)huawei.com> --- src/uadk.h | 5 +++-- src/uadk_cipher.c | 7 +++++-- src/uadk_dh.c | 7 +++++-- src/uadk_digest.c | 7 +++++-- src/uadk_pkey.c | 7 +++++-- src/uadk_rsa.c | 7 +++++-- 6 files changed, 28 insertions(+), 12 deletions(-) diff --git a/src/uadk.h b/src/uadk.h index d210c3d..267da89 100644 --- a/src/uadk.h +++ b/src/uadk.h @@ -21,8 +21,9 @@ #include <uadk/wd_sched.h> #include "uadk_utils.h" -#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) -#define ENV_STRING_LEN 256 +#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) +#define ENV_STRING_LEN 256 +#define ENGINE_RECV_MAX_CNT 60000000 enum { KUNPENG920, diff --git a/src/uadk_cipher.c b/src/uadk_cipher.c index 0b99753..5993386 100644 --- a/src/uadk_cipher.c +++ b/src/uadk_cipher.c @@ -503,6 +503,7 @@ static int sched_single_poll_policy(handle_t h_sched_ctx, static int uadk_e_cipher_poll(void *ctx) { struct cipher_priv_ctx *priv = (struct cipher_priv_ctx *) ctx; + __u64 rx_cnt = 0; __u32 recv = 0; /* Poll one packet currently */ int expt = 1; @@ -519,9 +520,11 @@ static int uadk_e_cipher_poll(void *ctx) return 0; else if (ret < 0 && ret != -EAGAIN) return ret; - } while (ret == -EAGAIN); + } while (ret == -EAGAIN && (rx_cnt++ < ENGINE_RECV_MAX_CNT)); - return ret; + fprintf(stderr, "failed to recv msg: timeout!\n"); + + return -ETIMEDOUT; } static int uadk_e_cipher_env_poll(void *ctx) diff --git a/src/uadk_dh.c b/src/uadk_dh.c index 75c9e5a..018a3b6 100644 --- a/src/uadk_dh.c +++ b/src/uadk_dh.c @@ -204,6 +204,7 @@ static __u32 dh_pick_next_ctx(handle_t sched_ctx, static int uadk_e_dh_poll(void *ctx) { + __u64 rx_cnt = 0; __u32 recv = 0; int expect = 1; int idx = 1; @@ -215,9 +216,11 @@ static int uadk_e_dh_poll(void *ctx) return UADK_E_POLL_SUCCESS; else if (ret < 0 && ret != -EAGAIN) return ret; - } while (ret == -EAGAIN); + } while (ret == -EAGAIN && (rx_cnt++ < ENGINE_RECV_MAX_CNT)); - return ret; + fprintf(stderr, "failed to recv msg: timeout!\n"); + + return -ETIMEDOUT; } static void uadk_e_dh_cb(void *req_t) diff --git a/src/uadk_digest.c b/src/uadk_digest.c index 70f51d3..0fb2cfb 100644 --- a/src/uadk_digest.c +++ b/src/uadk_digest.c @@ -331,6 +331,7 @@ static int sched_single_poll_policy(handle_t h_sched_ctx, static int uadk_e_digest_poll(void *ctx) { + __u64 rx_cnt = 0; __u32 recv = 0; int expt = 1; int ret = 0; @@ -341,9 +342,11 @@ static int uadk_e_digest_poll(void *ctx) return 0; else if (ret < 0 && ret != -EAGAIN) return ret; - } while (ret == -EAGAIN); + } while (ret == -EAGAIN && (rx_cnt++ < ENGINE_RECV_MAX_CNT)); - return ret; + fprintf(stderr, "failed to recv msg: timeout!\n"); + + return -ETIMEDOUT; } static int uadk_e_digest_env_poll(void *ctx) diff --git a/src/uadk_pkey.c b/src/uadk_pkey.c index 5300327..019d4fd 100644 --- a/src/uadk_pkey.c +++ b/src/uadk_pkey.c @@ -113,6 +113,7 @@ void uadk_ecc_cb(void *req_t) static int uadk_ecc_poll(void *ctx) { unsigned int recv = 0; + __u64 rx_cnt = 0; int expt = 1; int ret; @@ -122,9 +123,11 @@ static int uadk_ecc_poll(void *ctx) return 0; else if (ret < 0 && ret != -EAGAIN) return ret; - } while (ret == -EAGAIN); + } while (ret == -EAGAIN && (rx_cnt++ < ENGINE_RECV_MAX_CNT)); - return ret; + fprintf(stderr, "failed to recv msg: timeout!\n"); + + return -ETIMEDOUT; } /* make resource configure static */ diff --git a/src/uadk_rsa.c b/src/uadk_rsa.c index 1579539..42c3506 100644 --- a/src/uadk_rsa.c +++ b/src/uadk_rsa.c @@ -658,6 +658,7 @@ static int rsa_poll_policy(handle_t h_sched_ctx, __u32 expect, __u32 *count) static int uadk_e_rsa_poll(void *ctx) { + __u64 rx_cnt = 0; __u32 recv = 0; int expt = 1; int ret; @@ -668,9 +669,11 @@ static int uadk_e_rsa_poll(void *ctx) return UADK_E_POLL_SUCCESS; else if (ret < 0 && ret != -EAGAIN) return ret; - } while (ret == -EAGAIN); + } while (ret == -EAGAIN && (rx_cnt++ < ENGINE_RECV_MAX_CNT)); - return ret; + fprintf(stderr, "failed to recv msg: timeout!\n"); + + return -ETIMEDOUT; } static struct rsa_res_config rsa_res_config = { -- 2.30.0

1 0

[PATCH 0/5] uadk - add wd_comp_init2
by Yang Shen 07 Jul '22

07 Jul '22

The current uadk initialization process is: 1.Call wd_request_ctx() to request ctxs from devices. 2.Call wd_sched_rr_alloc() to create a sched(or some other scheduler alloc function if exits). 3.Initialize the sched. 4.Call wd_<alg>_init() with ctx_config and sched. Logic is reasonable. But in practice, the step of `wd_ request_ Ctx() ` and `wd_ sched_ rr_alloc() ` are very tedious. This makes it difficult for users to use the interface. One of the main reasons for this is that uadk has made a lot of configurations in the scheduler in order to provide users with better performance. Based on this consideration, the current uadk requires the user to arrange the division of hardware resources according to the device topology during initialization. Therefore, as a high-level interface, this scheme can provide customized scheme configuration for users with deep needs. All algorithm initialization interfaces have the same input parameters and behavioral logic. The pre-processing of the wd_<alg>_init is actually the configuration of `struct wd_ctx_config` and `struct wd_sched`. Therefore, the next thing to be done is to use limited and easy-to-use input parameters to describe users' requirements on the two input parameters, ensuring that the functions of the new interface init2 are the same as those of init. For ease of description, v1 is used to refer to the existing interface, and v2 is used to refer to the layer of encapsulation. At present, at least 4 parameters are required to meet the user configuration requirements with the V1 interface function remains unchanged. @device_list: The available uacce device list. Users can get it by wd_get_accel_list(). @numa_bitmask: The bitmask provided by libnuma. Users can use this parameter to control requesting ctxs devices in the bind NUMA scenario. @ctx_nums: The requested ctx number for each numa node. Due to users may have different requirements for different types of ctx numbers, needs a two-dimensional array as input. @sched_type: Scheduling type the user wants to use. What's more, some users want uadk to provide the default value about input parameters for some performance insensitive scenes. C code has no way to. Yang Shen (5): uadk - support algorithms initialization reentry protect uadk - support return error number as pointer uadk - mv some function to header file uadk/comp - add wd_comp_init2 uadk/docs - support a simple interface for initialization Makefile.am | 4 +- docs/wd_alg_init2.md | 176 +++++++++++++++++++++++++++ include/wd.h | 54 ++++++++- include/wd_alg_common.h | 24 ++++ include/wd_comp.h | 28 +++++ include/wd_util.h | 57 +++++++++ wd.c | 97 ++++++++++++--- wd_aead.c | 33 +++-- wd_cipher.c | 35 ++++-- wd_comp.c | 259 ++++++++++++++++++++++++++++++++++++++-- wd_dh.c | 34 ++++-- wd_digest.c | 33 +++-- wd_ecc.c | 33 +++-- wd_rsa.c | 33 +++-- wd_util.c | 75 +++++++++++- 15 files changed, 870 insertions(+), 105 deletions(-) create mode 100644 docs/wd_alg_init2.md -- 2.24.0

2 7

帐户验证!
by openeuler.orgIT 06 Jul '22

06 Jul '22

1 0

通知acc@openeuler.org
by noreply 02 Jul '22

02 Jul '22

1 0

[PATCH 0/2] uadk/sched - make a numa node map
by Yang Shen 01 Jul '22

01 Jul '22

Now the uadk scheduler will search the nearest node when call session_sched_init(). But this is a repetitive work because session_sched_init() is called by wd_xxx_create_sess(). So we can build an cpu-device numa node map. Update the map when a new numa node has been enabled with a device. And then, session_sched_init() can quick get the right device id from a certain cpu. Yang Shen (2): uadk/sched - adjust some functions location and rename for tag uadk/sched - add mapping from cpus to devices wd_sched.c | 312 ++++++++++++++++++++++++++++------------------------- 1 file changed, 163 insertions(+), 149 deletions(-) -- 2.24.0

1 2

[PATCH v3] uacce: Handle parent device removal or parent driver module rmmod
by Zhangfei Gao 01 Jul '22

01 Jul '22

From: Jean-Philippe Brucker <jean-philippe(a)linaro.org> The uacce driver must deal with a possible removal of the parent device or parent driver module rmmod at any time. Although uacce_remove(), called on device removal and on driver unbind, prevents future use of the uacce fops by removing the cdev, fops that were called before that point may still be running. Serialize uacce_fops_open() and uacce_remove() with uacce->mutex. Serialize other fops against uacce_remove() with q->mutex. Since we need to protect uacce_fops_poll() which gets called on the fast path, replace uacce->queues_lock with q->mutex to improve scalability. The other fops are only used during setup. uacce_queue_is_valid(), checked under q->mutex or uacce->mutex, denotes whether uacce_remove() has disabled all queues. If that is the case, don't go any further since the parent device is being removed and uacce->ops should not be called anymore. Reported-by: Yang Shen <shenyang39(a)huawei.com> Signed-off-by: Zhangfei Gao <zhangfei.gao(a)linaro.org> Signed-off-by: Jean-Philippe Brucker <jean-philippe(a)linaro.org> --- v3: Drop patch 1, which set cdev owner with parent driver onwer, to block parent driver rmmod. Suggested from Greg, this behavior introduce layering violation. Only keep patch 2, which can cover both parent moduel rmmod and parent device removal. drivers/misc/uacce/uacce.c | 133 ++++++++++++++++++++++++------------- include/linux/uacce.h | 6 +- 2 files changed, 91 insertions(+), 48 deletions(-) diff --git a/drivers/misc/uacce/uacce.c b/drivers/misc/uacce/uacce.c index 281c54003edc..b70a013139c7 100644 --- a/drivers/misc/uacce/uacce.c +++ b/drivers/misc/uacce/uacce.c @@ -9,43 +9,38 @@ static struct class *uacce_class; static dev_t uacce_devt; -static DEFINE_MUTEX(uacce_mutex); static DEFINE_XARRAY_ALLOC(uacce_xa); -static int uacce_start_queue(struct uacce_queue *q) +/* + * If the parent driver or the device disappears, the queue state is invalid and + * ops are not usable anymore. + */ +static bool uacce_queue_is_valid(struct uacce_queue *q) { - int ret = 0; + return q->state == UACCE_Q_INIT || q->state == UACCE_Q_STARTED; +} - mutex_lock(&uacce_mutex); +static int uacce_start_queue(struct uacce_queue *q) +{ + int ret; - if (q->state != UACCE_Q_INIT) { - ret = -EINVAL; - goto out_with_lock; - } + if (q->state != UACCE_Q_INIT) + return -EINVAL; if (q->uacce->ops->start_queue) { ret = q->uacce->ops->start_queue(q); if (ret < 0) - goto out_with_lock; + return ret; } q->state = UACCE_Q_STARTED; - -out_with_lock: - mutex_unlock(&uacce_mutex); - - return ret; + return 0; } static int uacce_put_queue(struct uacce_queue *q) { struct uacce_device *uacce = q->uacce; - mutex_lock(&uacce_mutex); - - if (q->state == UACCE_Q_ZOMBIE) - goto out; - if ((q->state == UACCE_Q_STARTED) && uacce->ops->stop_queue) uacce->ops->stop_queue(q); @@ -54,8 +49,6 @@ static int uacce_put_queue(struct uacce_queue *q) uacce->ops->put_queue(q); q->state = UACCE_Q_ZOMBIE; -out: - mutex_unlock(&uacce_mutex); return 0; } @@ -65,20 +58,36 @@ static long uacce_fops_unl_ioctl(struct file *filep, { struct uacce_queue *q = filep->private_data; struct uacce_device *uacce = q->uacce; + long ret = -ENXIO; + + /* + * uacce->ops->ioctl() may take the mmap_lock when copying arg to/from + * user. Avoid a circular lock dependency with uacce_fops_mmap(), which + * gets called with mmap_lock held, by taking uacce->mutex instead of + * q->mutex. Doing this in uacce_fops_mmap() is not possible because + * uacce_fops_open() calls iommu_sva_bind_device(), which takes + * mmap_lock, while holding uacce->mutex. + */ + mutex_lock(&uacce->mutex); + if (!uacce_queue_is_valid(q)) + goto out_unlock; switch (cmd) { case UACCE_CMD_START_Q: - return uacce_start_queue(q); - + ret = uacce_start_queue(q); + break; case UACCE_CMD_PUT_Q: - return uacce_put_queue(q); - + ret = uacce_put_queue(q); + break; default: - if (!uacce->ops->ioctl) - return -EINVAL; - - return uacce->ops->ioctl(q, cmd, arg); + if (uacce->ops->ioctl) + ret = uacce->ops->ioctl(q, cmd, arg); + else + ret = -EINVAL; } +out_unlock: + mutex_unlock(&uacce->mutex); + return ret; } #ifdef CONFIG_COMPAT @@ -136,6 +145,13 @@ static int uacce_fops_open(struct inode *inode, struct file *filep) if (!q) return -ENOMEM; + mutex_lock(&uacce->mutex); + + if (!uacce->parent) { + ret = -EINVAL; + goto out_with_mem; + } + ret = uacce_bind_queue(uacce, q); if (ret) goto out_with_mem; @@ -152,10 +168,9 @@ static int uacce_fops_open(struct inode *inode, struct file *filep) filep->private_data = q; uacce->inode = inode; q->state = UACCE_Q_INIT; - - mutex_lock(&uacce->queues_lock); + mutex_init(&q->mutex); list_add(&q->list, &uacce->queues); - mutex_unlock(&uacce->queues_lock); + mutex_unlock(&uacce->mutex); return 0; @@ -163,18 +178,20 @@ static int uacce_fops_open(struct inode *inode, struct file *filep) uacce_unbind_queue(q); out_with_mem: kfree(q); + mutex_unlock(&uacce->mutex); return ret; } static int uacce_fops_release(struct inode *inode, struct file *filep) { struct uacce_queue *q = filep->private_data; + struct uacce_device *uacce = q->uacce; - mutex_lock(&q->uacce->queues_lock); - list_del(&q->list); - mutex_unlock(&q->uacce->queues_lock); + mutex_lock(&uacce->mutex); uacce_put_queue(q); uacce_unbind_queue(q); + list_del(&q->list); + mutex_unlock(&uacce->mutex); kfree(q); return 0; @@ -217,10 +234,9 @@ static int uacce_fops_mmap(struct file *filep, struct vm_area_struct *vma) vma->vm_private_data = q; qfr->type = type; - mutex_lock(&uacce_mutex); - - if (q->state != UACCE_Q_INIT && q->state != UACCE_Q_STARTED) { - ret = -EINVAL; + mutex_lock(&q->mutex); + if (!uacce_queue_is_valid(q)) { + ret = -ENXIO; goto out_with_lock; } @@ -248,12 +264,12 @@ static int uacce_fops_mmap(struct file *filep, struct vm_area_struct *vma) } q->qfrs[type] = qfr; - mutex_unlock(&uacce_mutex); + mutex_unlock(&q->mutex); return ret; out_with_lock: - mutex_unlock(&uacce_mutex); + mutex_unlock(&q->mutex); kfree(qfr); return ret; } @@ -262,12 +278,20 @@ static __poll_t uacce_fops_poll(struct file *file, poll_table *wait) { struct uacce_queue *q = file->private_data; struct uacce_device *uacce = q->uacce; + __poll_t ret = 0; + + mutex_lock(&q->mutex); + if (!uacce_queue_is_valid(q)) + goto out_unlock; poll_wait(file, &q->wait, wait); + if (uacce->ops->is_q_updated && uacce->ops->is_q_updated(q)) - return EPOLLIN | EPOLLRDNORM; + ret = EPOLLIN | EPOLLRDNORM; - return 0; +out_unlock: + mutex_unlock(&q->mutex); + return ret; } static const struct file_operations uacce_fops = { @@ -450,7 +474,7 @@ struct uacce_device *uacce_alloc(struct device *parent, goto err_with_uacce; INIT_LIST_HEAD(&uacce->queues); - mutex_init(&uacce->queues_lock); + mutex_init(&uacce->mutex); device_initialize(&uacce->dev); uacce->dev.devt = MKDEV(MAJOR(uacce_devt), uacce->dev_id); uacce->dev.class = uacce_class; @@ -507,13 +531,23 @@ void uacce_remove(struct uacce_device *uacce) if (uacce->inode) unmap_mapping_range(uacce->inode->i_mapping, 0, 0, 1); + /* + * uacce_fops_open() may be running concurrently, even after we remove + * the cdev. Holding uacce->mutex ensures that open() does not obtain a + * removed uacce device. + */ + mutex_lock(&uacce->mutex); /* ensure no open queue remains */ - mutex_lock(&uacce->queues_lock); list_for_each_entry_safe(q, next_q, &uacce->queues, list) { + /* + * Taking q->mutex ensures that fops do not use the defunct + * uacce->ops after the queue is disabled. + */ + mutex_lock(&q->mutex); uacce_put_queue(q); + mutex_unlock(&q->mutex); uacce_unbind_queue(q); } - mutex_unlock(&uacce->queues_lock); /* disable sva now since no opened queues */ uacce_disable_sva(uacce); @@ -521,6 +555,13 @@ void uacce_remove(struct uacce_device *uacce) if (uacce->cdev) cdev_device_del(uacce->cdev, &uacce->dev); xa_erase(&uacce_xa, uacce->dev_id); + /* + * uacce exists as long as there are open fds, but ops will be freed + * now. Ensure that bugs cause NULL deref rather than use-after-free. + */ + uacce->ops = NULL; + uacce->parent = NULL; + mutex_unlock(&uacce->mutex); put_device(&uacce->dev); } EXPORT_SYMBOL_GPL(uacce_remove); diff --git a/include/linux/uacce.h b/include/linux/uacce.h index 48e319f40275..9ce88c28b0a8 100644 --- a/include/linux/uacce.h +++ b/include/linux/uacce.h @@ -70,6 +70,7 @@ enum uacce_q_state { * @wait: wait queue head * @list: index into uacce queues list * @qfrs: pointer of qfr regions + * @mutex: protects queue state * @state: queue state machine * @pasid: pasid associated to the mm * @handle: iommu_sva handle returned by iommu_sva_bind_device() @@ -80,6 +81,7 @@ struct uacce_queue { wait_queue_head_t wait; struct list_head list; struct uacce_qfile_region *qfrs[UACCE_MAX_REGION]; + struct mutex mutex; enum uacce_q_state state; u32 pasid; struct iommu_sva *handle; @@ -97,9 +99,9 @@ struct uacce_queue { * @dev_id: id of the uacce device * @cdev: cdev of the uacce * @dev: dev of the uacce + * @mutex: protects uacce operation * @priv: private pointer of the uacce * @queues: list of queues - * @queues_lock: lock for queues list * @inode: core vfs */ struct uacce_device { @@ -113,9 +115,9 @@ struct uacce_device { u32 dev_id; struct cdev *cdev; struct device dev; + struct mutex mutex; void *priv; struct list_head queues; - struct mutex queues_lock; struct inode *inode; }; -- 2.36.1

1 0