- Acc - mailweb.openeuler.org

[PATCH 00/23] uadk and uadk_tool bugfix
by Qi Tao 13 Jun '25

13 Jun '25

*** BLURB HERE *** Chenghai Huang (6): uadk：remove the data consumption modifying when copying cache output uadk: add the function of parsing the cached tail packets uadk: add the function of parsing the cached tail packets in v1 uadk: fix the windowbits setting issue in zlibwrapper uadk: fix gzip windowbits and level setting issue in zlibwrapper uadk: fix repeated initialization issue Hao Fang (1): uadk_tool: hpre: fix for x448/x25519 perf test failed Longfang Liu (2): uadk: removal of the Shared Memory Interface for Queues uadk_tools: Removal of the Shared Memory Interface for Queues Qi Tao (1): uadk: fix code cleanup issues Super User (1): uadk: the uadk_tool is compiled using the private OpenSSL header file. Zhiqi Song (1): uadk/hpre: remove redundant null operation Zhushuai Yin (3): Uadk: Fix the DAE memory leak issue. uadk: The MAX MIN and rehash functions are added to the hash agg. Uadk: fix hash agg review comments lizhi (8): uadk: adapt ECDH for high-performance cores uadk_tool: adapt ecdh benchmark with secp256r1 curve uadk/hpre: fix up severel parameter type mismatch issues uadk: check hardware type to config hpcore field uadk_tool: update ECC curve in uadk_tool benchmark uadk:fix some cleanup issues uadk: fix potential null pointer dereference uadk_tool: Fix x448 bug caused by wrong nid. Makefile.am | 2 + drv/hisi_comp.c | 54 +-- drv/hisi_comp_huf.c | 248 ++++++++++++++ drv/hisi_comp_huf.h | 19 ++ drv/hisi_dae.c | 244 ++++++++++++-- drv/hisi_hpre.c | 108 +++++- drv/hisi_qm_udrv.h | 3 +- include/drv/wd_agg_drv.h | 10 +- include/drv/wd_ecc_drv.h | 10 + include/wd_agg.h | 9 +- include/wd_ecc_curve.h | 27 ++ include/wd_util.h | 2 +- include/wd_zlibwrapper.h | 2 +- uadk_tool/benchmark/hpre_uadk_benchmark.c | 58 ++-- uadk_tool/benchmark/hpre_wd_benchmark.c | 32 +- uadk_tool/benchmark/sec_soft_benchmark.c | 20 +- uadk_tool/benchmark/uadk_benchmark.h | 2 +- uadk_tool/test/comp_lib.h | 2 +- v1/drv/hisi_zip_huf.c | 248 ++++++++++++++ v1/drv/hisi_zip_huf.h | 19 ++ v1/drv/hisi_zip_udrv.c | 52 ++- v1/test/hisi_hpre_test/hpre_test_tools.c | 392 ---------------------- v1/test/hisi_zip_test_sgl/wd_sched_sgl.c | 310 +++++++++-------- v1/test/test_mm/test_wd_mem.c | 8 +- v1/test/wd_sched.c | 247 +++++++------- v1/uacce.h | 1 - v1/wd.c | 57 +--- v1/wd.h | 2 - wd.c | 4 +- wd_aead.c | 8 +- wd_agg.c | 47 +-- wd_cipher.c | 16 +- wd_comp.c | 6 +- wd_dh.c | 4 +- wd_digest.c | 7 +- wd_ecc.c | 70 +++- wd_rsa.c | 5 +- wd_util.c | 6 +- wd_zlibwrapper.c | 48 +-- 39 files changed, 1473 insertions(+), 936 deletions(-) create mode 100644 drv/hisi_comp_huf.c create mode 100644 drv/hisi_comp_huf.h create mode 100644 v1/drv/hisi_zip_huf.c create mode 100644 v1/drv/hisi_zip_huf.h -- 2.33.0

1 23

【2025/6/11 AccLIb SIG 双周例会 11:00 - 12:00】
by openEuler conference 10 Jun '25

10 Jun '25

您好！ sig-AccLib 邀请您参加 2025-06-11 11:00 召开的Zoom会议会议主题：【2025/6/11 AccLIb SIG 双周例会 11:00 - 12:00】会议内容：会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet 会议链接：https://us06web.zoom.us/j/86361583097?pwd=PIujIKJR3JaIEl4pg3LAdclauyjQyy.1 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! sig-AccLib invites you to attend the Zoom conference will be held at 2025-06-11 11:00, The subject of the conference is 【2025/6/11 AccLIb SIG 双周例会 11:00 - 12:00】 Summary: 会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet You can join the meeting at https://us06web.zoom.us/j/86361583097?pwd=PIujIKJR3JaIEl4pg3LAdclauyjQyy.1 Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings More information: https://www.openeuler.org/en/

1 0

[PATCH 00/10] uadk_engine/provider: fixed static code alarms and code format issues.
by Qi Tao 05 Jun '25

05 Jun '25

From: JiangShui Yang <yangjiangshui(a)h-partners.com> Chenghai Huang (4): uadk_provider: fix some possible issues where empty memory is accessed uadk_provider: adjust the code style uadk_provider: fix the static variable problem in the case of multiple concurrency uadk_provider: add parameter validation to avoid memory errors Hao Fang (6): uadk_engine/provider: tinyfixes for code format alignment uadk_engine/provider: add blank line before return uadk_engine/provider: remove the variable redundant initialization uadk_provider: add const keyword uadk_provider: remove unused parameters uadk_engine:sm2: use UADK_E_INVAL replace -1 src/uadk_aead.c | 2 +- src/uadk_cipher.c | 13 +-- src/uadk_dh.c | 11 +-- src/uadk_digest.c | 24 +++--- src/uadk_ec.c | 6 ++ src/uadk_ecx.c | 4 +- src/uadk_engine_init.c | 1 + src/uadk_pkey.c | 9 ++- src/uadk_pkey.h | 3 +- src/uadk_prov_aead.c | 3 +- src/uadk_prov_bio.c | 18 ++--- src/uadk_prov_cipher.c | 46 ++++++----- src/uadk_prov_der_writer.c | 96 +++++++++++----------- src/uadk_prov_der_writer.h | 8 +- src/uadk_prov_dh.c | 89 +++++++++++---------- src/uadk_prov_digest.c | 23 +++--- src/uadk_prov_ec_kmgmt.c | 3 +- src/uadk_prov_ecdh_exch.c | 21 +++-- src/uadk_prov_ecdsa.c | 8 +- src/uadk_prov_ecx.c | 53 ++++++------ src/uadk_prov_ffc.c | 160 ++++++++++++++++++++----------------- src/uadk_prov_ffc.h | 4 +- src/uadk_prov_hmac.c | 20 ++--- src/uadk_prov_init.c | 10 +-- src/uadk_prov_packet.c | 48 ++++++----- src/uadk_prov_packet.h | 52 ++++++------ src/uadk_prov_pkey.c | 63 ++++++++------- src/uadk_prov_rsa.c | 82 +++++++++---------- src/uadk_prov_sm2.c | 135 ++++++++++++++++--------------- src/uadk_rsa.c | 31 +++---- src/uadk_sm2.c | 59 ++++++++------ 31 files changed, 594 insertions(+), 511 deletions(-) -- 2.33.0

1 10

[PATCH 1/2] uadk_engine/hpre: fix memory leaks in sm2
by Qi Tao 28 May '25

28 May '25

From: lizhi <lizhi206(a)huawei.com> fix memory leak and avoid possible double-free risk in sm2. Signed-off-by: lizhi <lizhi206(a)huawei.com> --- src/uadk_sm2.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/src/uadk_sm2.c b/src/uadk_sm2.c index 170d320..7737292 100644 --- a/src/uadk_sm2.c +++ b/src/uadk_sm2.c @@ -407,9 +407,12 @@ static int sign_bin_to_ber(EC_KEY *ec, struct wd_dtb *r, struct wd_dtb *s, if (sltmp < 0) { fprintf(stderr, "failed to i2d_ECDSA_SIG\n"); ret = -EINVAL; - goto free_s; + /* bs and br set to e_sig, use unified interface to prevent double release. */ + goto free_sig; } + *siglen = (size_t)sltmp; + ECDSA_SIG_free(e_sig); return 0; free_s: @@ -417,7 +420,6 @@ free_s: free_r: BN_clear_free(br); free_sig: - ECDSA_SIG_set0(e_sig, NULL, NULL); ECDSA_SIG_free(e_sig); return ret; -- 2.33.0

1 1

[PATCH] uadk_prov: fix some cleancode issues
by Qi Tao 28 May '25

28 May '25

Fixed the following issues: 1. Remove redundant function return value check. 2. Modify inappropriate exception printing. 3. There is no need to check the return value of async_wake_job(). Only need to check req->state to determine if there is an exception in packet reception. Therefore, add "(void)" before the function call. 4. Free the allocated memory before returning an error. 5. The callback function's parameter list does not match the function pointer definition. Signed-off-by: Qi Tao <taoqi10(a)huawei.com> --- src/uadk_aead.c | 4 ++-- src/uadk_async.c | 2 +- src/uadk_cipher.c | 6 +----- src/uadk_dh.c | 2 +- src/uadk_digest.c | 2 +- src/uadk_pkey.c | 2 +- src/uadk_prov_aead.c | 4 ++-- src/uadk_prov_cipher.c | 4 ++-- src/uadk_prov_dh.c | 2 +- src/uadk_prov_digest.c | 6 +++--- src/uadk_prov_hmac.c | 6 +++--- src/uadk_prov_init.c | 4 +++- src/uadk_prov_pkey.c | 2 +- src/uadk_prov_rsa.c | 2 +- src/uadk_rsa.c | 2 +- 15 files changed, 24 insertions(+), 26 deletions(-) diff --git a/src/uadk_aead.c b/src/uadk_aead.c index 1da7753..69223f1 100644 --- a/src/uadk_aead.c +++ b/src/uadk_aead.c @@ -469,7 +469,7 @@ static int do_aead_sync_inner(struct aead_priv_ctx *priv, unsigned char *out, priv->req.state = 0; ret = wd_do_aead_sync(priv->sess, &priv->req); if (unlikely(ret < 0 || priv->req.state)) { - fprintf(stderr, "do aead task failed, msg state: %d, ret: %d, state: %u!\n", + fprintf(stderr, "do aead task failed, msg state: %u, ret: %d, state: %u!\n", state, ret, priv->req.state); return RET_FAIL; } @@ -547,7 +547,7 @@ static void *uadk_e_aead_cb(struct wd_aead_req *req, void *data) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } return NULL; diff --git a/src/uadk_async.c b/src/uadk_async.c index 7536bd5..0f92dde 100644 --- a/src/uadk_async.c +++ b/src/uadk_async.c @@ -329,7 +329,7 @@ static void *async_poll_process_func(void *args) if (!poll_queue.is_recv && op->job) { op->done = 1; op->ret = ret; - async_wake_job(op->job); + (void)async_wake_job(op->job); async_free_poll_task(idx, 0); } } diff --git a/src/uadk_cipher.c b/src/uadk_cipher.c index 457e90d..95af193 100644 --- a/src/uadk_cipher.c +++ b/src/uadk_cipher.c @@ -242,10 +242,6 @@ static int uadk_e_cipher_soft_work(EVP_CIPHER_CTX *ctx, unsigned char *out, */ if (!priv->update_iv) { iv = EVP_CIPHER_CTX_iv_noconst(ctx); - if (unlikely(iv == NULL)) { - fprintf(stderr, "get openssl software iv failed.\n"); - return 0; - } memcpy(iv, priv->iv, EVP_CIPHER_CTX_iv_length(ctx)); priv->update_iv = true; } @@ -567,7 +563,7 @@ static void *uadk_e_cipher_cb(struct wd_cipher_req *req, void *data) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } return NULL; diff --git a/src/uadk_dh.c b/src/uadk_dh.c index 2f2c1cf..011bf56 100644 --- a/src/uadk_dh.c +++ b/src/uadk_dh.c @@ -259,7 +259,7 @@ static void uadk_e_dh_cb(void *req_t) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } } diff --git a/src/uadk_digest.c b/src/uadk_digest.c index 6827f98..0b4e8c5 100644 --- a/src/uadk_digest.c +++ b/src/uadk_digest.c @@ -802,7 +802,7 @@ static void *uadk_e_digest_cb(void *data) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } return NULL; diff --git a/src/uadk_pkey.c b/src/uadk_pkey.c index 1f8234b..e05c7d0 100644 --- a/src/uadk_pkey.c +++ b/src/uadk_pkey.c @@ -101,7 +101,7 @@ void uadk_e_ecc_cb(void *req_t) op->done = 1; op->ret = 0; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } } diff --git a/src/uadk_prov_aead.c b/src/uadk_prov_aead.c index 54e0115..dbbd844 100644 --- a/src/uadk_prov_aead.c +++ b/src/uadk_prov_aead.c @@ -390,7 +390,7 @@ static void *uadk_prov_aead_cb(struct wd_aead_req *req, void *data) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } return NULL; @@ -800,7 +800,7 @@ static int uadk_get_aead_info(struct aead_priv_ctx *priv) } if (unlikely(i == aead_counts)) { - fprintf(stderr, "failed to setup the private ctx.\n"); + fprintf(stderr, "failed to get aead info.\n"); return UADK_AEAD_FAIL; } diff --git a/src/uadk_prov_cipher.c b/src/uadk_prov_cipher.c index db4a957..dfe6666 100644 --- a/src/uadk_prov_cipher.c +++ b/src/uadk_prov_cipher.c @@ -395,7 +395,7 @@ static int uadk_get_cipher_info(struct cipher_priv_ctx *priv) } } - fprintf(stderr, "failed to setup the private ctx.\n"); + fprintf(stderr, "failed to get cipher info.\n"); return UADK_P_FAIL; } @@ -455,7 +455,7 @@ static void async_cb(struct wd_cipher_req *req, void *data) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } } diff --git a/src/uadk_prov_dh.c b/src/uadk_prov_dh.c index c75c386..1cb4b45 100644 --- a/src/uadk_prov_dh.c +++ b/src/uadk_prov_dh.c @@ -581,7 +581,7 @@ static void uadk_prov_dh_cb(void *req_t) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } } diff --git a/src/uadk_prov_digest.c b/src/uadk_prov_digest.c index 7a5bfff..a13d075 100644 --- a/src/uadk_prov_digest.c +++ b/src/uadk_prov_digest.c @@ -303,7 +303,7 @@ static int uadk_get_digest_info(struct digest_priv_ctx *priv) } if (unlikely(i == digest_counts)) { - fprintf(stderr, "failed to setup the private ctx.\n"); + fprintf(stderr, "failed to digest info.\n"); return UADK_DIGEST_FAIL; } @@ -534,7 +534,7 @@ soft_update: return uadk_digest_soft_update(priv, data, data_len); } -static void uadk_async_cb(struct wd_digest_req *req, void *data) +static void uadk_async_cb(struct wd_digest_req *req) { struct uadk_e_cb_info *digest_cb_param; struct wd_digest_req *req_origin; @@ -550,7 +550,7 @@ static void uadk_async_cb(struct wd_digest_req *req, void *data) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } } diff --git a/src/uadk_prov_hmac.c b/src/uadk_prov_hmac.c index db49612..6bf7947 100644 --- a/src/uadk_prov_hmac.c +++ b/src/uadk_prov_hmac.c @@ -361,7 +361,7 @@ static int uadk_get_hmac_info(struct hmac_priv_ctx *priv) } } - fprintf(stderr, "failed to setup the private ctx, algname = %s.\n", priv->alg_name); + fprintf(stderr, "failed to get hmac info, algname = %s.\n", priv->alg_name); return UADK_HMAC_FAIL; } @@ -529,7 +529,7 @@ soft_init: return uadk_hmac_soft_init(priv); } -static void uadk_hmac_async_cb(struct wd_digest_req *req, void *data) +static void uadk_hmac_async_cb(struct wd_digest_req *req) { struct uadk_e_cb_info *hmac_cb_param; struct wd_digest_req *req_origin; @@ -545,7 +545,7 @@ static void uadk_hmac_async_cb(struct wd_digest_req *req, void *data) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } } diff --git a/src/uadk_prov_init.c b/src/uadk_prov_init.c index c29500b..20f7068 100644 --- a/src/uadk_prov_init.c +++ b/src/uadk_prov_init.c @@ -470,8 +470,10 @@ int OSSL_provider_init(const OSSL_CORE_HANDLE *handle, ctx->libctx = (OSSL_LIB_CTX *)c_get_libctx(handle); ret = uadk_prov_ctx_set_core_bio_method(ctx); - if (!ret) + if (!ret) { + OPENSSL_free(ctx); return UADK_P_FAIL; + } ret = async_module_init(); if (!ret) diff --git a/src/uadk_prov_pkey.c b/src/uadk_prov_pkey.c index ca853ae..ac4541c 100644 --- a/src/uadk_prov_pkey.c +++ b/src/uadk_prov_pkey.c @@ -362,7 +362,7 @@ void uadk_prov_ecc_cb(void *req_t) ecc_async_op->done = 1; ecc_async_op->ret = 0; async_free_poll_task(ecc_async_op->idx, 1); - async_wake_job(ecc_async_op->job); + (void)async_wake_job(ecc_async_op->job); } } diff --git a/src/uadk_prov_rsa.c b/src/uadk_prov_rsa.c index b9713cc..4351514 100644 --- a/src/uadk_prov_rsa.c +++ b/src/uadk_prov_rsa.c @@ -1275,7 +1275,7 @@ static void uadk_e_rsa_cb(void *req_t) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } } diff --git a/src/uadk_rsa.c b/src/uadk_rsa.c index 76678a4..1755374 100644 --- a/src/uadk_rsa.c +++ b/src/uadk_rsa.c @@ -1117,7 +1117,7 @@ static void uadk_e_rsa_cb(void *req_t) if (op && op->job && !op->done) { op->done = 1; async_free_poll_task(op->idx, 1); - async_wake_job(op->job); + (void)async_wake_job(op->job); } } -- 2.33.0

1 0

【2025/5/28 AccLIb SIG 双周例会 11:00 - 12:00】
by openEuler conference 27 May '25

27 May '25

您好！ sig-AccLib 邀请您参加 2025-05-28 11:00 召开的Zoom会议会议主题：【2025/5/28 AccLIb SIG 双周例会 11:00 - 12:00】会议内容：会议链接： linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet 会议链接：https://us06web.zoom.us/j/82863167501?pwd=tvnic0aWCxFSVFObmeHHXbN3gh2a8m.1 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! sig-AccLib invites you to attend the Zoom conference will be held at 2025-05-28 11:00, The subject of the conference is 【2025/5/28 AccLIb SIG 双周例会 11:00 - 12:00】 Summary: 会议链接： linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet You can join the meeting at https://us06web.zoom.us/j/82863167501?pwd=tvnic0aWCxFSVFObmeHHXbN3gh2a8m.1 Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings More information: https://www.openeuler.org/en/

1 0

[PATCH 00/22] uadk_engine: some bugfix and cleanup
by Qi Tao 27 May '25

27 May '25

From: JiangShui Yang <yangjiangshui(a)h-partners.com> Chenghai Huang (5): uadk_provider: adjust the threshold for SHA384 to use Digest to Calculated key uadk_provider: fix dh memery leak issue in free params uadk_provider: fix rsa memery leak issue in bn_ctx_new uadk_provider: rectify the soft switch contidition of RAS uadk_provider: fix some compilation warning issues Qi Tao (6): uadk_prov: fix some cleancode issues uadk_prov: the query returns the v3 alg array to handle exception uadk_prov: modify the location of creating soft ctx uadk_prov: fix memory leak error uadk_engine: fix cleancode issues uadk_prov: default to turn on the soft switch Weili Qian (3): uadk_engine: fix possible double free issues uadk_engine/ecdh: remove unnecessary call uadk_ecc_set_public_key uadk_provider/sm2: free all applied memory when freectx Zhiqi Song (2): uadk_engine/cipher: fixup the method of obtaining IV in init uadk_engine/hpre: fixup some param release problem Zhushuai Yin (1): uadk_provider:fix dh insufficient performance lizhi (5): uadk_engine/ecdh: Pad the high bits of short shared keys with 0 uadk_provider/ecdh: Pad the high bits of short shared keys with 0 uadk_provider/ecx: fix memory leak of ecx_key uadk_provider/sm2: Fix the race conditions caused by atomic variables uadk_provider/sm2: fix a bug of switching soft compute src/uadk_aead.c | 4 +- src/uadk_async.c | 2 +- src/uadk_cipher.c | 8 +- src/uadk_dh.c | 4 +- src/uadk_digest.c | 5 +- src/uadk_ec.c | 52 +++--- src/uadk_pkey.c | 2 +- src/uadk_prov.h | 4 +- src/uadk_prov_aead.c | 43 +++-- src/uadk_prov_cipher.c | 38 +++-- src/uadk_prov_dh.c | 123 ++++++++------ src/uadk_prov_digest.c | 38 +++-- src/uadk_prov_ecdh_exch.c | 9 +- src/uadk_prov_ecx.c | 29 ++-- src/uadk_prov_ffc.c | 170 +++++++++++++++++--- src/uadk_prov_ffc.h | 38 ++++- src/uadk_prov_hmac.c | 8 +- src/uadk_prov_init.c | 36 +++-- src/uadk_prov_pkey.c | 2 +- src/uadk_prov_rsa.c | 138 ++++++++-------- src/uadk_prov_sm2.c | 331 ++++++++++++++++++++------------------ src/uadk_rsa.c | 2 +- src/uadk_sm2.c | 10 +- 23 files changed, 678 insertions(+), 418 deletions(-) -- 2.33.0

1 22

[PATCH] uadk_provider: fix the async issue when receive fail
by Qi Tao 26 May '25

26 May '25

From: Chenghai Huang <huangchenghai2(a)huawei.com> The corresponding sending thread can be precisely woken up only after the poll thread receiving succeeds. When a hardware error occurs, asynchronous receiving will reture an abnormal value. In this case, the corresponding sending thread cannot be sure to woken up. Therefore, the counting method is used to wake up the corresponding thread whose number of failures exceeds the threshold. Signed-off-by: Chenghai Huang <huangchenghai2(a)huawei.com> --- src/uadk_async.c | 10 ++++++++++ src/uadk_async.h | 2 ++ 2 files changed, 12 insertions(+) diff --git a/src/uadk_async.c b/src/uadk_async.c index 0f92dde..8adc964 100644 --- a/src/uadk_async.c +++ b/src/uadk_async.c @@ -327,6 +327,16 @@ static void *async_poll_process_func(void *args) idx = op->idx; ret = async_recv_func[task->type](task->ctx); if (!poll_queue.is_recv && op->job) { + /* + * Wake up the op->job whose number of polling + * failures exceeds the threshold. + */ + if (ret && ret != -ETIMEDOUT && op->retry++ < ASYNC_MAX_RETRY_CNT) { + (void)sem_post(&poll_queue.full_sem); + continue; + } + + op->retry = 0; op->done = 1; op->ret = ret; (void) async_wake_job(op->job); diff --git a/src/uadk_async.h b/src/uadk_async.h index 8de3c3c..23f017e 100644 --- a/src/uadk_async.h +++ b/src/uadk_async.h @@ -23,6 +23,7 @@ #include <openssl/async.h> #define ASYNC_QUEUE_TASK_NUM 1024 +#define ASYNC_MAX_RETRY_CNT 1000 #define UADK_E_SUCCESS 1 #define UADK_E_FAIL 0 #define DO_SYNC 1 @@ -32,6 +33,7 @@ struct async_op { int done; int idx; int ret; + int retry; }; struct uadk_e_cb_info { -- 2.33.0

1 0

[PATCH] sanity_test: fix blk hw test
by Qi Tao 22 May '25

22 May '25

From: Hao Fang <fanghao11(a)huawei.com> Signed-off-by: Hao Fang <fanghao11(a)huawei.com> --- test/sanity_test.sh | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/test/sanity_test.sh b/test/sanity_test.sh index 1c4b631..acc44a9 100755 --- a/test/sanity_test.sh +++ b/test/sanity_test.sh @@ -87,10 +87,10 @@ hw_blk_deflate() case $3 in "gzip") ${RM} -f /tmp/gzip_list.bin - run_cmd uadk_tool test --m zip --in $1 --out $2 $@ + run_cmd uadk_tool test --m zip --stream --in $1 --out $2 $@ ;; "zlib") - run_cmd uadk_tool test --m zip --alg 1 --in $1 --out $2 $@ + run_cmd uadk_tool test --m zip --stream --alg 1 --in $1 --out $2 $@ ;; *) echo "Unsupported algorithm type: $3" @@ -104,10 +104,10 @@ hw_blk_inflate() { case $3 in "gzip") - run_cmd uadk_tool test --m zip --inf --in $1 --out $2 $@ + run_cmd uadk_tool test --m zip --stream --inf --in $1 --out $2 $@ ;; "zlib") - run_cmd uadk_tool test --m zip --alg 1 --inf --in $1 --out $2 $@ + run_cmd uadk_tool test --m zip --stream --alg 1 --inf --in $1 --out $2 $@ ;; *) echo "Unsupported algorithm type: $3" @@ -365,9 +365,9 @@ run_zip_test_v2() WD_COMP_EPOLL_EN=1 hw_dfl_hw_ifl /tmp/syslog WD_COMP_EPOLL_EN=0 hw_dfl_hw_ifl /tmp/syslog # test without environment variables - #run_cmd uadk_tool test --m zip --blksize 8192 --size 81920 --loop 1000 --self + #run_cmd uadk_tool test --m zip --stream --blksize 8192 --size 81920 --loop 1000 --self # test with environment variables - #run_cmd uadk_tool test --m zip --blksize 8192 --size 81920 --loop 1000 --self --env + #run_cmd uadk_tool test --m zip --stream --blksize 8192 --size 81920 --loop 1000 --self --env } # Accept more paraterms -- 2.33.0

1 0

[PATCH] uadk_provider: fix the async issue when receive fail
by Qi Tao 21 May '25

21 May '25

From: Chenghai Huang <huangchenghai2(a)huawei.com> When a hardware error occurs, asynchronous receiving will reture an abnormal value. In this case, the corresponding sending thread cannot be sure to woken up. Therefore, the counting method is used to wake up the corresponding thread whose number of failures exceeds the threshold. Signed-off-by: Chenghai Huang <huangchenghai2(a)huawei.com> --- src/uadk_async.c | 8 ++++++++ src/uadk_async.h | 2 ++ 2 files changed, 10 insertions(+) diff --git a/src/uadk_async.c b/src/uadk_async.c index 0f92dde..947b4ee 100644 --- a/src/uadk_async.c +++ b/src/uadk_async.c @@ -326,6 +326,14 @@ static void *async_poll_process_func(void *args) op = task->op; idx = op->idx; ret = async_recv_func[task->type](task->ctx); + if (ret && !poll_queue.is_recv) { + if (op->retry++ < ASYNC_MAX_RETRY_CNT) { + (void)sem_post(&poll_queue.full_sem); + continue; + } + op->retry = 0; + } + if (!poll_queue.is_recv && op->job) { op->done = 1; op->ret = ret; diff --git a/src/uadk_async.h b/src/uadk_async.h index 8de3c3c..23f017e 100644 --- a/src/uadk_async.h +++ b/src/uadk_async.h @@ -23,6 +23,7 @@ #include <openssl/async.h> #define ASYNC_QUEUE_TASK_NUM 1024 +#define ASYNC_MAX_RETRY_CNT 1000 #define UADK_E_SUCCESS 1 #define UADK_E_FAIL 0 #define DO_SYNC 1 @@ -32,6 +33,7 @@ struct async_op { int done; int idx; int ret; + int retry; }; struct uadk_e_cb_info { -- 2.33.0

1 0

[PATCH] test: remove --ilist --olist commands for list_loader.py
by Qi Tao 15 May '25

15 May '25

From: Hao Fang <fanghao11(a)huawei.com> list_loader.py is for software comp test, but hw comp test had remove the --list commands, the loopback test will be failed, so list_loader.py also do the same thing. Signed-off-by: Hao Fang <fanghao11(a)huawei.com> --- test/list_loader.py | 71 ++++++++++++++------------------------------- test/sanity_test.sh | 4 +-- 2 files changed, 23 insertions(+), 52 deletions(-) diff --git a/test/list_loader.py b/test/list_loader.py index d5e1578..7373a88 100755 --- a/test/list_loader.py +++ b/test/list_loader.py @@ -13,22 +13,16 @@ class listcontent(object): self.ofile_nm = ofile self.ifile = open(ifile, "rb") self.ofile = open(ofile, "wb") - # addr: 8 bytes, size: 8 bytes, next: 8 bytes - #self.entry_sz = 0x18; - self.entry_addr = 0; - self.entry_size = 0; - self.entry_next = 0; def __del__(self): self.ifile.close() self.ofile.close() - def deflate(self, olist, blk_sz): + def deflate(self, blk_sz): ifile_sz = os.path.getsize(self.ifile_nm) count = (ifile_sz + int(blk_sz) - 1) / int(blk_sz) # Create array data = np.ndarray(count * 3, dtype=np.uint64) - entries = data.reshape(-1, 3) i = 0 while i < count: # Each block of data is stored in temporary file that is used @@ -44,33 +38,21 @@ class listcontent(object): os.system("cat %s > %s" % (of.name, self.ofile_nm)) else: os.system("cat %s >> %s" % (of.name, self.ofile_nm)) - # entries[i][0] should be the address of output buffer. - # But the output is file now, not buffer. So fill it with - # any non-zero value. - entries[i][0] = 1 - entries[i][1] = os.path.getsize(of.name) - if i == count - 1: - entries[i][2] = 0 - else: - entries[i][2] = 1 i += 1 os.remove(f.name) os.remove(of.name) - entries.tofile(olist) - # Read block data from ifile by ilist. And inflate each block data. - def inflate(self, ilist): - self.ilist_nm = ilist - data = np.fromfile(self.ilist_nm, dtype=np.uint64) - # Each entry contains addr, size and next fields - # Convert data array into the two dimensional array - entries = data.reshape(-1, 3) + # Read block data from ifile. And inflate each block data. + def inflate(self): + ifile_sz = os.path.getsize(self.ifile_nm) i = 0 - while i < entries.shape[0]: + while True: # Each block of data is stored in temporary file that is used # by gunzip. f = tempfile.NamedTemporaryFile(delete=False) - blk = self.ifile.read(entries[i][1]) + blk = self.ifile.read(ifile_sz) + if not blk: + break f.write(blk) f.close() if not i: @@ -78,42 +60,31 @@ class listcontent(object): else: os.system("gunzip < %s >> %s" % (f.name, self.ofile_nm)) os.remove(f.name) - if not entries[i][2]: - break i += 1 -def sw_deflate(ifile, ofile, olist, blk_sz): +def sw_deflate(ifile, ofile, blk_sz): dfl = listcontent(ifile, ofile) - dfl.deflate(olist, blk_sz) + dfl.deflate(blk_sz) -def sw_inflate(ifile, ofile, ilist): +def sw_inflate(ifile, ofile): ifl = listcontent(ifile, ofile) - ifl.inflate(ilist) + ifl.inflate() def main(argv): - ilist = '' - olist = '' ifile = '' ofile = '' blk_sz = 0 try: - opts, args = getopt.getopt(argv, "hb:", ["ilist=","olist=","in=","out="]) + opts, args = getopt.getopt(argv, "hb:", ["in=","out="]) for opt, arg in opts: if opt in ("-h"): print('Software deflate command:') - print(' list_loader -b <block size> --in <file> --out <file> --olist <file>') + print(' list_loader -b <block size> --in <file> --out <file>') print('Software inflate command:') - print(' list_loader --in <file> --out <file> --ilist <file>') + print(' list_loader --in <file> --out <file>') sys.exit() elif opt in ("-b"): blk_sz = arg - elif opt in ("--ilist"): - if not os.path.isfile(arg): - print("File does not exist:", arg) - sys.exit(1) - ilist = arg - elif opt in ("--olist"): - olist = arg elif opt in ("--in"): if not os.path.isfile(arg): print("File does not exist:", arg) @@ -122,18 +93,18 @@ def main(argv): elif opt in ("--out"): ofile = arg except getopt.GetoptError: - # Compress source to destination file and output list file + # Compress source to destination file print('Software deflate command:') - print(' list_loader -b <block size> --in <file> --out <file> --olist <file>') - # Decompress source to destination file with input list file + print(' list_loader -b <block size> --in <file> --out <file>') + # Decompress source to destination file print('Software inflate command:') - print(' list_loader --in <file> --out <file> --ilist <file>') + print(' list_loader --in <file> --out <file>') sys.exit(2) if blk_sz: - sw_deflate(ifile, ofile, olist, blk_sz) + sw_deflate(ifile, ofile, blk_sz) else: - sw_inflate(ifile, ofile, ilist) + sw_inflate(ifile, ofile) if __name__ == "__main__": main(sys.argv[1:]) diff --git a/test/sanity_test.sh b/test/sanity_test.sh index 9ffb137..1c4b631 100755 --- a/test/sanity_test.sh +++ b/test/sanity_test.sh @@ -157,7 +157,7 @@ sw_blk_deflate() case $3 in "gzip") ${RM} -f /tmp/gzip_list.bin - python test/list_loader.py --in $1 --out $2 --olist /tmp/gzip_list.bin -b $4 + python test/list_loader.py --in $1 --out $2 -b $4 ;; *) echo "Unsupported algorithm type: $3" @@ -171,7 +171,7 @@ sw_blk_inflate() { case $3 in "gzip") - python test/list_loader.py --in $1 --out $2 --ilist /tmp/gzip_list.bin + python test/list_loader.py --in $1 --out $2 ;; *) echo "Unsupported algorithm type: $3" -- 2.33.0

1 0

[PATCH] [acc/uadk_tool] Modify malloc for queue init
by Qi Tao 08 May '25

08 May '25

From: Junchong Pan <panjunchong(a)h-partners.com> Queue is malloc but not init, is_poll is set but poll interface is not invoked, may case eq overflow. Signed-off-by: Junchong Pan <panjunchong(a)h-partners.com> Signed-off-by: Qi Tao <taoqi10(a)huawei.com> --- uadk_tool/benchmark/hpre_wd_benchmark.c | 2 +- uadk_tool/benchmark/sec_wd_benchmark.c | 2 +- uadk_tool/benchmark/zip_wd_benchmark.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) diff --git a/uadk_tool/benchmark/hpre_wd_benchmark.c b/uadk_tool/benchmark/hpre_wd_benchmark.c index 40f7f0b..60ed3bd 100644 --- a/uadk_tool/benchmark/hpre_wd_benchmark.c +++ b/uadk_tool/benchmark/hpre_wd_benchmark.c @@ -447,7 +447,7 @@ static int init_hpre_wd_queue(struct acc_option *options) } for (i = 0; i < g_thread_num; i++) { - g_thread_queue.bd_res[i].queue = malloc(sizeof(struct wd_queue)); + g_thread_queue.bd_res[i].queue = calloc(1, sizeof(struct wd_queue)); g_thread_queue.bd_res[i].queue->capa.alg = options->algclass; // 0 is ENC, 1 is DEC g_thread_queue.bd_res[i].queue->capa.priv.direction = options->optype; diff --git a/uadk_tool/benchmark/sec_wd_benchmark.c b/uadk_tool/benchmark/sec_wd_benchmark.c index e022dcb..f066a5b 100644 --- a/uadk_tool/benchmark/sec_wd_benchmark.c +++ b/uadk_tool/benchmark/sec_wd_benchmark.c @@ -610,7 +610,7 @@ static int init_wd_queue(struct acc_option *options) } for (i = 0; i < g_thread_num; i++) { - g_thread_queue.bd_res[i].queue = malloc(sizeof(struct wd_queue)); + g_thread_queue.bd_res[i].queue = calloc(1, sizeof(struct wd_queue)); memset(g_thread_queue.bd_res[i].queue, 0, sizeof(struct wd_queue)); g_thread_queue.bd_res[i].queue->capa.alg = options->algclass; // 0 is ENC, 1 is DEC diff --git a/uadk_tool/benchmark/zip_wd_benchmark.c b/uadk_tool/benchmark/zip_wd_benchmark.c index 8ad3e96..ab9a894 100644 --- a/uadk_tool/benchmark/zip_wd_benchmark.c +++ b/uadk_tool/benchmark/zip_wd_benchmark.c @@ -311,7 +311,7 @@ static int init_zip_wd_queue(struct acc_option *options) } for (i = 0; i < g_thread_num; i++) { - g_thread_queue.bd_res[i].queue = malloc(sizeof(struct wd_queue)); + g_thread_queue.bd_res[i].queue = calloc(1, sizeof(struct wd_queue)); g_thread_queue.bd_res[i].queue->capa.alg = options->algclass; // 0 is compress, 1 is decompress g_thread_queue.bd_res[i].queue->capa.priv.direction = op_type; -- 2.33.0

1 1

2025/4/30 AccLIb SIG 双周例会 11:00 - 12:00
by openEuler conference 29 Apr '25

29 Apr '25

您好！ sig-AccLib 邀请您参加 2025-04-30 11:00 召开的Zoom会议会议主题：2025/4/30 AccLIb SIG 双周例会 11:00 - 12:00 会议内容：会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet 会议链接：https://us06web.zoom.us/j/87831643076?pwd=Xx3vcsnmCOBXKdtaKVecCJjSRboVNl.1 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! sig-AccLib invites you to attend the Zoom conference will be held at 2025-04-30 11:00, The subject of the conference is 2025/4/30 AccLIb SIG 双周例会 11:00 - 12:00 Summary: 会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet You can join the meeting at https://us06web.zoom.us/j/87831643076?pwd=Xx3vcsnmCOBXKdtaKVecCJjSRboVNl.1 Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings More information: https://www.openeuler.org/en/

1 0

[PATCH] sanity_test: replace zip_sva_perf by uadk_tool test --m zip
by Qi Tao 25 Apr '25

25 Apr '25

From: Hao Fang <fanghao11(a)huawei.com> uadk_tool test --m zip supports the most of function tests except for one command --ilist, so the test file size set to 8MB for block mode. The --ilist cmd has been removed because a complex user protocol was constructed only for large file compression in block mode. Signed-off-by: Hao Fang <fanghao11(a)huawei.com> Signed-off-by: Qi Tao <taoqi10(a)huawei.com> --- test/sanity_test.sh | 62 +++++++++++++++++++------------------- uadk_tool/test/comp_main.c | 5 --- 2 files changed, 31 insertions(+), 36 deletions(-) diff --git a/test/sanity_test.sh b/test/sanity_test.sh index 591f8e2d..a1f2b234 100755 --- a/test/sanity_test.sh +++ b/test/sanity_test.sh @@ -88,10 +88,10 @@ hw_blk_deflate() case $3 in "gzip") ${RM} -f /tmp/gzip_list.bin - run_cmd zip_sva_perf --in $1 --out $2 --olist /tmp/gzip_list.bin $@ + run_cmd uadk_tool test --m zip --in $1 --out $2 $@ ;; "zlib") - run_cmd zip_sva_perf -z --in $1 --out $2 --olist /tmp/zlib_list.bin $@ + run_cmd uadk_tool test --m zip --alg 1 --in $1 --out $2 $@ ;; *) echo "Unsupported algorithm type: $3" @@ -105,10 +105,10 @@ hw_blk_inflate() { case $3 in "gzip") - run_cmd zip_sva_perf -d --in $1 --out $2 --ilist /tmp/gzip_list.bin $@ + run_cmd uadk_tool test --m zip --inf --in $1 --out $2 $@ ;; "zlib") - run_cmd zip_sva_perf -z -d --in $1 --out $2 --ilist /tmp/zlib_list.bin $@ + run_cmd uadk_tool test --m zip --alg 1 --inf --in $1 --out $2 $@ ;; *) echo "Unsupported algorithm type: $3" @@ -122,10 +122,10 @@ hw_strm_deflate() { case $3 in "gzip") - run_cmd zip_sva_perf -S --in $1 --out $2 $@ + run_cmd uadk_tool test --m zip --stream --in $1 --out $2 $@ ;; "zlib") - run_cmd zip_sva_perf -z -S --in $1 --out $2 $@ + run_cmd uadk_tool test --m zip --stream --alg 1 --in $1 --out $2 $@ ;; *) echo "Unsupported algorithm type: $3" @@ -139,10 +139,10 @@ hw_strm_inflate() { case $3 in "gzip") - run_cmd zip_sva_perf -S -d --in $1 --out $2 $@ + run_cmd uadk_tool test --m zip --stream --inf --in $1 --out $2 $@ ;; "zlib") - run_cmd zip_sva_perf -z -S -d --in $1 --out $2 $@ + run_cmd uadk_tool test --m zip --stream --inf --in $1 --out $2 $@ ;; *) echo "Unsupported algorithm type: $3" @@ -234,12 +234,12 @@ hw_dfl_sw_ifl() prepare_src_file random 1 md5sum origin > ori.md5 - hw_blk_deflate origin /tmp/ori.gz gzip -b 8192 --env + hw_blk_deflate origin /tmp/ori.gz gzip --blksize 8192 --env sw_blk_inflate /tmp/ori.gz origin gzip md5sum -c ori.md5 ${RM} -f /tmp/ori.gz - hw_strm_deflate origin /tmp/ori.gz gzip -b 8192 --env + hw_strm_deflate origin /tmp/ori.gz gzip --blksize 8192 --env sw_strm_inflate /tmp/ori.gz origin gzip md5sum -c ori.md5 @@ -249,7 +249,7 @@ hw_dfl_sw_ifl() md5sum origin > ori.md5 ${RM} -f /tmp/ori.gz - hw_strm_deflate origin /tmp/ori.gz gzip -b 8192 --env + hw_strm_deflate origin /tmp/ori.gz gzip --blksize 8192 --env sw_strm_inflate /tmp/ori.gz origin gzip md5sum -c ori.md5 @@ -259,13 +259,13 @@ hw_dfl_sw_ifl() prepare_src_file $1 md5sum origin > ori.md5 - hw_blk_deflate origin /tmp/ori.gz gzip -b 8192 --env + hw_blk_deflate origin /tmp/ori.gz gzip --blksize 8192 --env sw_blk_inflate /tmp/ori.gz origin gzip md5sum -c ori.md5 # This case fails. ${RM} -f /tmp/ori.gz - hw_strm_deflate origin /tmp/ori.gz gzip -b 8192 --env + hw_strm_deflate origin /tmp/ori.gz gzip --blksize 8192 --env sw_strm_inflate /tmp/ori.gz origin gzip md5sum -c ori.md5 } @@ -282,13 +282,13 @@ sw_dfl_hw_ifl() # Only gzip compress and hardware decompress sw_blk_deflate origin /tmp/ori.gz gzip 8192 - hw_blk_inflate /tmp/ori.gz origin gzip -b 8192 + hw_blk_inflate /tmp/ori.gz origin gzip --blksize 8192 md5sum -c ori.md5 - hw_blk_inflate /tmp/ori.gz origin gzip -b 8192 --env + hw_blk_inflate /tmp/ori.gz origin gzip --blksize 8192 --env md5sum -c ori.md5 sw_strm_deflate origin /tmp/ori.gz gzip 8192 - hw_strm_inflate /tmp/ori.gz origin gzip -b 8192 --env + hw_strm_inflate /tmp/ori.gz origin gzip --blksize 8192 --env md5sum -c ori.md5 # Generate random data with 500MB size @@ -298,7 +298,7 @@ sw_dfl_hw_ifl() ${RM} -f /tmp/ori.gz sw_strm_deflate origin /tmp/ori.gz gzip 8192 - hw_strm_inflate /tmp/ori.gz origin gzip -b 8192 --env + hw_strm_inflate /tmp/ori.gz origin gzip --blksize 8192 --env md5sum -c ori.md5 # Use existed text file. It's not in alignment. @@ -308,11 +308,11 @@ sw_dfl_hw_ifl() md5sum origin > ori.md5 #sw_blk_deflate origin /tmp/ori.gz gzip 8192 - #hw_blk_inflate /tmp/ori.gz origin gzip -b 8192 --env + #hw_blk_inflate /tmp/ori.gz origin gzip --blksize 8192 --env #md5sum -c ori.md5 sw_strm_deflate origin /tmp/ori.gz gzip 8192 - hw_strm_inflate /tmp/ori.gz origin gzip -b 8192 --env + hw_strm_inflate /tmp/ori.gz origin gzip --blksize 8192 --env md5sum -c ori.md5 } @@ -326,13 +326,13 @@ hw_dfl_hw_ifl() prepare_src_file random 1 md5sum origin > ori.md5 - hw_blk_deflate origin /tmp/ori.gz gzip -b 8192 - hw_blk_inflate /tmp/ori.gz origin gzip -b 8192 + hw_blk_deflate origin /tmp/ori.gz gzip --blksize 8192 + hw_blk_inflate /tmp/ori.gz origin gzip --blksize 8192 md5sum -c ori.md5 ${RM} -f /tmp/ori.gz - hw_strm_deflate origin /tmp/ori.gz gzip -b 8192 - hw_strm_inflate /tmp/ori.gz origin gzip -b 8192 + hw_strm_deflate origin /tmp/ori.gz gzip --blksize 8192 + hw_strm_inflate /tmp/ori.gz origin gzip --blksize 8192 md5sum -c ori.md5 # Use existed text file. It's not in alignment. @@ -341,13 +341,13 @@ hw_dfl_hw_ifl() prepare_src_file $1 md5sum origin > ori.md5 - #hw_blk_deflate origin /tmp/ori.gz gzip -b 8192 - #hw_blk_inflate /tmp/ori.gz origin gzip -b 8192 + #hw_blk_deflate origin /tmp/ori.gz gzip --blksize 8192 + #hw_blk_inflate /tmp/ori.gz origin gzip --blksize 8192 #md5sum -c ori.md5 ${RM} -f /tmp/ori.gz - hw_strm_deflate origin /tmp/ori.gz gzip -b 8192 - hw_strm_inflate /tmp/ori.gz origin gzip -b 8192 + hw_strm_deflate origin /tmp/ori.gz gzip --blksize 8192 + hw_strm_inflate /tmp/ori.gz origin gzip --blksize 8192 md5sum -c ori.md5 } @@ -358,17 +358,17 @@ run_zip_test_v2() export WD_COMP_ASYNC_POLL_EN=1 export WD_COMP_ASYNC_POLL_NUM="4@0" # test without environment variables - # limit test file in 64MB + # limit test file in 8MB rm -fr /tmp/syslog - dd if=/var/log/syslog of=/tmp/syslog bs=1M count=16 >& /dev/null + dd if=/var/log/syslog of=/tmp/syslog bs=1M count=8 >& /dev/null sw_dfl_hw_ifl /tmp/syslog hw_dfl_sw_ifl /tmp/syslog WD_COMP_EPOLL_EN=1 hw_dfl_hw_ifl /tmp/syslog WD_COMP_EPOLL_EN=0 hw_dfl_hw_ifl /tmp/syslog # test without environment variables - #run_cmd zip_sva_perf -b 8192 -s 81920 -l 1000 --self + #run_cmd uadk_tool test --m zip --blksize 8192 --size 81920 --loop 1000 --self # test with environment variables - #run_cmd zip_sva_perf -b 8192 -s 81920 -l 1000 --self --env + #run_cmd uadk_tool test --m zip --blksize 8192 --size 81920 --loop 1000 --self --env } # Accept more paraterms diff --git a/uadk_tool/test/comp_main.c b/uadk_tool/test/comp_main.c index 3bdfea62..77d5decc 100644 --- a/uadk_tool/test/comp_main.c +++ b/uadk_tool/test/comp_main.c @@ -674,11 +674,6 @@ int store_file(struct hizip_test_info *info, char *model) size_t sum = 0; ssize_t file_sz = 0; - if (!opts->is_stream) { - COMP_TST_PRT("Invalid, file need stream mode!\n"); - return -EINVAL; - } - p = tdata->out_list; file_sz = write(opts->fd_out, p->addr, p->size); if (file_sz < p->size) -- 2.33.0

1 0

[PATCH 01/29] uadk: fix for aead soft compute
by Qi Tao 24 Apr '25

24 Apr '25

From: Wenkai Lin <linwenkai6(a)hisilicon.com> When soft computing is required, an invalid BD is used to ensure the integrity of the sending and receiving process, it is more efficient. Signed-off-by: Wenkai Lin <linwenkai6(a)hisilicon.com> Signed-off-by: Qi Tao <taoqi10(a)huawei.com> --- drv/hisi_sec.c | 49 ++++++++++++------------------------------------- 1 file changed, 12 insertions(+), 37 deletions(-) diff --git a/drv/hisi_sec.c b/drv/hisi_sec.c index 7f4ae109..5990565a 100644 --- a/drv/hisi_sec.c +++ b/drv/hisi_sec.c @@ -54,6 +54,8 @@ #define SEC_AI_GEN_OFFSET_V3 2 #define SEC_SEQ_OFFSET_V3 6 #define SEC_AUTH_MASK_V3 0xFFFFFFFC +#define SEC_BD3_IVLD_OFFSET 0x4 +#define SEC_BD2_IVLD_OFFSET 0x7 #define SEC_SGL_MODE_MASK_V3 0x4800 #define SEC_PBUFF_MODE_MASK_V3 0x800 @@ -2440,7 +2442,7 @@ static int gcm_do_soft_mac(struct wd_aead_msg *msg) msg->result = WD_SUCCESS; - return WD_SOFT_COMPUTING; + return WD_SUCCESS; } static int fill_stream_bd2(struct wd_aead_msg *msg, struct hisi_sec_sqe *sqe) @@ -2460,6 +2462,8 @@ static int fill_stream_bd2(struct wd_aead_msg *msg, struct hisi_sec_sqe *sqe) if (msg->cmode == WD_CIPHER_GCM) { gcm_auth_ivin(msg); ret = gcm_do_soft_mac(msg); + /* Make the bd invalid to avoid recalculation of the data. */ + sqe->iv_tls_ld = 0x1 << SEC_BD2_IVLD_OFFSET; } break; default: @@ -2577,12 +2581,8 @@ static int hisi_sec_aead_send(struct wd_alg_driver *drv, handle_t ctx, void *wd_ fill_aead_bd2_addr(msg, &sqe); ret = fill_stream_bd2(msg, &sqe); - if (ret == WD_SOFT_COMPUTING) { - ret = 0; - goto put_sgl; - } else if (unlikely(ret)) { + if (unlikely(ret)) goto put_sgl; - } hisi_set_msg_id(h_qp, &msg->tag); sqe.type2.tag = (__u16)msg->tag; @@ -2661,24 +2661,6 @@ static void parse_aead_bd2(struct hisi_qp *qp, struct hisi_sec_sqe *sqe, dump_sec_msg(temp_msg, "aead"); } -static bool soft_compute_check(struct hisi_qp *qp, struct wd_aead_msg *msg) -{ - /* Asynchronous mode does not use the sent message, so ignores it */ - if (qp->q_info.qp_mode == CTX_MODE_ASYNC) - return false; - /* - * For aead gcm stream mode, due to some hardware limitations, - * the final message was not sent to hardware if the qm is - * not higher than v3 version or the input length of the - * message is 0, the software calculation has been executed. - */ - if (msg->msg_state == AEAD_MSG_END && msg->cmode == WD_CIPHER_GCM && - (qp->q_info.hw_type <= HISI_QM_API_VER3_BASE || !msg->in_bytes)) - return true; - - return false; -} - static int hisi_sec_aead_recv(struct wd_alg_driver *drv, handle_t ctx, void *wd_msg) { handle_t h_qp = (handle_t)wd_ctx_get_priv(ctx); @@ -2687,9 +2669,6 @@ static int hisi_sec_aead_recv(struct wd_alg_driver *drv, handle_t ctx, void *wd_ __u16 count = 0; int ret; - if (soft_compute_check((struct hisi_qp *)h_qp, recv_msg)) - return 0; - ret = hisi_qm_recv(h_qp, &sqe, 1, &count); if (ret < 0) return ret; @@ -2886,10 +2865,13 @@ static int fill_stream_bd3(handle_t h_qp, struct wd_aead_msg *msg, struct hisi_s if (msg->cmode == WD_CIPHER_GCM) { gcm_auth_ivin(msg); /* Due to hardware limitations, software compute is required. */ - if (qp->q_info.hw_type <= HISI_QM_API_VER3_BASE || !msg->in_bytes) + if (qp->q_info.hw_type <= HISI_QM_API_VER3_BASE || !msg->in_bytes) { ret = gcm_do_soft_mac(msg); - else + /* Make the bd invalid to avoid recalculation of the data. */ + sqe->bd_param |= 0x1 << SEC_BD3_IVLD_OFFSET; + } else { fill_gcm_final_bd3(msg, sqe); + } } break; default: @@ -2982,12 +2964,8 @@ static int hisi_sec_aead_send_v3(struct wd_alg_driver *drv, handle_t ctx, void * fill_aead_bd3_addr(msg, &sqe); ret = fill_stream_bd3(h_qp, msg, &sqe); - if (ret == WD_SOFT_COMPUTING) { - ret = 0; - goto put_sgl; - } else if (unlikely(ret)) { + if (unlikely(ret)) goto put_sgl; - } hisi_set_msg_id(h_qp, &msg->tag); sqe.tag = msg->tag; @@ -3059,9 +3037,6 @@ static int hisi_sec_aead_recv_v3(struct wd_alg_driver *drv, handle_t ctx, void * __u16 count = 0; int ret; - if (soft_compute_check((struct hisi_qp *)h_qp, recv_msg)) - return 0; - ret = hisi_qm_recv(h_qp, &sqe, 1, &count); if (ret < 0) return ret; -- 2.33.0

1 28

[PATCH 00/17] uadk_provider: add switching software computation function
by Qi Tao 21 Apr '25

21 Apr '25

From: JiangShui Yang <yangjiangshui(a)h-partners.com> Chenghai Huang (4): uadk_provider: fix a memory leak issue in BIO_meth uadk_provider: add hmac alg for uadk_provider in openssl3.0 uadk_provider: add soft mac switching function for uadk_provider uadk_engine: fix a digest software switching issue in update Junchong Pan (1): uadk_engine - fix print format Qi Tao (2): uadk_provider: fix async packet reception timeout error uadk_prov: sec support switching to soft calculation Wenkai Lin (2): digest: fix for digest soft work digest: fix for digest soft init Zhiqi Song (1): uadk_provider: fix atomic compare parameter lizhi (7): uadk_provider: support switching to software in init uadk_provider/dh: add dh switching software computation function uadk_provider/rsa: add rsa switching software computation function uadk_provider/sm2: add sm2 switching software computation function uadk_provider/ecx: add x448 and x25519 switching software computation function uadk_provider/ecdsa: add ecdsa switching software computation function uadk_provider/ecdh: add ecdh switching software computation function src/Makefile.am | 3 +- src/uadk_async.h | 1 + src/uadk_digest.c | 108 ++-- src/uadk_ec.c | 2 +- src/uadk_prov.h | 11 + src/uadk_prov_aead.c | 340 ++++++++--- src/uadk_prov_cipher.c | 93 +++- src/uadk_prov_dh.c | 88 ++- src/uadk_prov_digest.c | 220 +++++--- src/uadk_prov_ec_kmgmt.c | 27 +- src/uadk_prov_ecdh_exch.c | 61 +- src/uadk_prov_ecdsa.c | 69 ++- src/uadk_prov_ecx.c | 242 +++++--- src/uadk_prov_hmac.c | 1115 +++++++++++++++++++++++++++++++++++++ src/uadk_prov_init.c | 171 ++++-- src/uadk_prov_pkey.c | 30 +- src/uadk_prov_pkey.h | 6 + src/uadk_prov_rsa.c | 182 +++--- src/uadk_prov_sm2.c | 241 +++++++- 19 files changed, 2526 insertions(+), 484 deletions(-) create mode 100644 src/uadk_prov_hmac.c -- 2.33.0

1 17

2025/4/16 AccLIb SIG 双周例会 11:00 - 12:00
by openEuler conference 15 Apr '25

15 Apr '25

您好！ sig-AccLib 邀请您参加 2025-04-16 11:00 召开的Zoom会议会议主题：2025/4/16 AccLIb SIG 双周例会 11:00 - 12:00 会议内容：会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet 会议链接：https://us06web.zoom.us/j/85069156271?pwd=tWrjUA5saZVqZ4PxRJX7XMjqZv4GNa.1 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! sig-AccLib invites you to attend the Zoom conference will be held at 2025-04-16 11:00, The subject of the conference is 2025/4/16 AccLIb SIG 双周例会 11:00 - 12:00 Summary: 会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet You can join the meeting at https://us06web.zoom.us/j/85069156271?pwd=tWrjUA5saZVqZ4PxRJX7XMjqZv4GNa.1 Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings More information: https://www.openeuler.org/en/

1 0

2025/4/2 AccLIb SIG 双周例会 11:00 - 12:00
by openEuler conference 01 Apr '25

01 Apr '25

您好！ sig-AccLib 邀请您参加 2025-04-02 11:00 召开的Zoom会议会议主题：2025/4/2 AccLIb SIG 双周例会 11:00 - 12:00 会议内容：会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet 会议链接：https://us06web.zoom.us/j/89619281074?pwd=U8YN8vo04bS9kKqwQntJjtamrbUeYY.1 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! sig-AccLib invites you to attend the Zoom conference will be held at 2025-04-02 11:00, The subject of the conference is 2025/4/2 AccLIb SIG 双周例会 11:00 - 12:00 Summary: 会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet You can join the meeting at https://us06web.zoom.us/j/89619281074?pwd=U8YN8vo04bS9kKqwQntJjtamrbUeYY.1 Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings More information: https://www.openeuler.org/en/

1 0

[PATCH 1/3] tests/qtest: Allow DSDT acpi tables to change
by Zhou Wang 31 Mar '25

31 Mar '25

From: caijian <caijian11(a)h-partners.com> List all DSDT files and allow them to change. Signed-of-by: caijian <caijian11(a)h-partners.com> --- tests/qtest/bios-tables-test-allowed-diff.h | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/tests/qtest/bios-tables-test-allowed-diff.h b/tests/qtest/bios-tables-test-allowed-diff.h index dfb8523c8b..e4a94bb8bd 100644 --- a/tests/qtest/bios-tables-test-allowed-diff.h +++ b/tests/qtest/bios-tables-test-allowed-diff.h @@ -1 +1,7 @@ /* List of comma-separated changed AML files to ignore */ +"tests/data/acpi/microvm/DSDT.pcie", +"tests/data/acpi/virt/DSDT", +"tests/data/acpi/virt/DSDT.acpihmatvirt", +"tests/data/acpi/virt/DSDT.memhp", +"tests/data/acpi/virt/DSDT.pxb", +"tests/data/acpi/virt/DSDT.topology", -- 2.43.0

1 5

【2025/3/19 AccLIb SIG 双周例会 11:00 - 12:00】
by openEuler conference 18 Mar '25

18 Mar '25

您好！ sig-AccLib 邀请您参加 2025-03-19 11:00 召开的Zoom会议会议主题：【2025/3/19 AccLIb SIG 双周例会 11:00 - 12:00】会议内容：会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet 会议链接：https://us06web.zoom.us/j/83002354706?pwd=X0OBGbq6auYKYZcvtSXaJqLNODbCfs.1 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! sig-AccLib invites you to attend the Zoom conference will be held at 2025-03-19 11:00, The subject of the conference is 【2025/3/19 AccLIb SIG 双周例会 11:00 - 12:00】 Summary: 会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet You can join the meeting at https://us06web.zoom.us/j/83002354706?pwd=X0OBGbq6auYKYZcvtSXaJqLNODbCfs.1 Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings More information: https://www.openeuler.org/en/

1 0

2025/3/5 AccLIb SIG 双周例会 11:00 - 12:00
by openEuler conference 04 Mar '25

04 Mar '25

您好！ sig-AccLib 邀请您参加 2025-03-05 11:00 召开的Zoom会议会议主题：2025/3/5 AccLIb SIG 双周例会 11:00 - 12:00 会议内容：会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meetings 会议链接：https://us06web.zoom.us/j/86433360382?pwd=sHbExElIDMtdjFUNtt2A4ZfbFqTssp.1 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! sig-AccLib invites you to attend the Zoom conference will be held at 2025-03-05 11:00, The subject of the conference is 2025/3/5 AccLIb SIG 双周例会 11:00 - 12:00 Summary: 会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meetings You can join the meeting at https://us06web.zoom.us/j/86433360382?pwd=sHbExElIDMtdjFUNtt2A4ZfbFqTssp.1 Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings More information: https://www.openeuler.org/en/

1 0

2025/2/19 AccLIb SIG 双周例会 11:00 - 12:00
by openEuler conference 18 Feb '25

18 Feb '25

您好！ sig-AccLib 邀请您参加 2025-02-19 11:00 召开的Zoom会议会议主题：2025/2/19 AccLIb SIG 双周例会 11:00 - 12:00 会议内容：会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet 会议链接：https://us06web.zoom.us/j/84927213524?pwd=QpZ05fkCFtfN0j2vRR0JN5jUsXaR7l.1 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! sig-AccLib invites you to attend the Zoom conference will be held at 2025-02-19 11:00, The subject of the conference is 2025/2/19 AccLIb SIG 双周例会 11:00 - 12:00 Summary: 会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet You can join the meeting at https://us06web.zoom.us/j/84927213524?pwd=QpZ05fkCFtfN0j2vRR0JN5jUsXaR7l.1 Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings More information: https://www.openeuler.org/en/

1 0

[PATCH] digest: fix for digest soft init
by Qi Tao 05 Feb '25

05 Feb '25

From: Wenkai Lin <linwenkai6(a)hisilicon.com> Ensure that soft_ctx is initialized before invoking software calculation and reduces repeated judgments for digest soft init. Signed-off-by: Wenkai Lin <linwenkai6(a)hisilicon.com> --- src/uadk_digest.c | 83 ++++++++++++++++++++--------------------------- 1 file changed, 36 insertions(+), 47 deletions(-) diff --git a/src/uadk_digest.c b/src/uadk_digest.c index ed68ee2..f7882e4 100644 --- a/src/uadk_digest.c +++ b/src/uadk_digest.c @@ -189,70 +189,56 @@ static EVP_MD *uadk_sha512; static int uadk_e_digests_soft_md(struct digest_priv_ctx *priv) { - int app_datasize = 0; + int app_datasize; - if (priv->soft_md) + if (priv->soft_md && priv->soft_ctx) return 1; switch (priv->e_nid) { - case NID_sm3: - priv->soft_md = EVP_sm3(); - break; case NID_md5: priv->soft_md = EVP_md5(); + /* OpenSSL 3.0 has no app_datasize, need set manually + * check crypto/evp/legacy_md5.c: md5_md as example + */ + app_datasize = EVP_MD_meth_get_app_datasize(priv->soft_md); + if (!app_datasize) + app_datasize = sizeof(EVP_MD *) + sizeof(MD5_CTX); + break; + case NID_sm3: + priv->soft_md = EVP_sm3(); + app_datasize = EVP_MD_meth_get_app_datasize(priv->soft_md); + if (!app_datasize) + app_datasize = sizeof(EVP_MD *) + sizeof(SM3_CTX); break; case NID_sha1: priv->soft_md = EVP_sha1(); + app_datasize = EVP_MD_meth_get_app_datasize(priv->soft_md); + if (!app_datasize) + app_datasize = sizeof(EVP_MD *) + sizeof(SHA_CTX); break; case NID_sha224: - priv->soft_md = EVP_sha224(); - break; case NID_sha256: - priv->soft_md = EVP_sha256(); + if (priv->e_nid == NID_sha224) + priv->soft_md = EVP_sha224(); + else + priv->soft_md = EVP_sha256(); + app_datasize = EVP_MD_meth_get_app_datasize(priv->soft_md); + if (!app_datasize) + app_datasize = sizeof(EVP_MD *) + sizeof(SHA256_CTX); break; case NID_sha384: - priv->soft_md = EVP_sha384(); - break; case NID_sha512: - priv->soft_md = EVP_sha512(); + if (priv->e_nid == NID_sha384) + priv->soft_md = EVP_sha384(); + else + priv->soft_md = EVP_sha512(); + app_datasize = EVP_MD_meth_get_app_datasize(priv->soft_md); + if (!app_datasize) + app_datasize = sizeof(EVP_MD *) + sizeof(SHA512_CTX); break; default: - break; - } - - if (unlikely(priv->soft_md == NULL)) + fprintf(stderr, "digest nid %d is invalid.\n", priv->e_nid); return 0; - - app_datasize = EVP_MD_meth_get_app_datasize(priv->soft_md); - if (app_datasize == 0) { - /* OpenSSL 3.0 has no app_datasize, need set manually - * check crypto/evp/legacy_md5.c: md5_md as example - */ - switch (priv->e_nid) { - case NID_sm3: - app_datasize = sizeof(EVP_MD *) + sizeof(SM3_CTX); - break; - case NID_md5: - app_datasize = sizeof(EVP_MD *) + sizeof(MD5_CTX); - break; - case NID_sha1: - app_datasize = sizeof(EVP_MD *) + sizeof(SHA_CTX); - break; - case NID_sha224: - app_datasize = sizeof(EVP_MD *) + sizeof(SHA256_CTX); - break; - case NID_sha256: - app_datasize = sizeof(EVP_MD *) + sizeof(SHA256_CTX); - break; - case NID_sha384: - app_datasize = sizeof(EVP_MD *) + sizeof(SHA512_CTX); - break; - case NID_sha512: - app_datasize = sizeof(EVP_MD *) + sizeof(SHA512_CTX); - break; - default: - break; - } } if (priv->soft_ctx == NULL) { @@ -262,8 +248,10 @@ static int uadk_e_digests_soft_md(struct digest_priv_ctx *priv) return 0; ctx->md_data = OPENSSL_malloc(app_datasize); - if (ctx->md_data == NULL) + if (ctx->md_data == NULL) { + EVP_MD_CTX_free(ctx); return 0; + } priv->soft_ctx = ctx; priv->app_datasize = app_datasize; @@ -1015,6 +1003,7 @@ static int uadk_e_digest_copy(EVP_MD_CTX *to, const EVP_MD_CTX *from) } if (t->soft_ctx) { + t->soft_md = NULL; t->soft_ctx = NULL; ret = digest_soft_init(t); if (!ret) -- 2.33.0

1 0

1.22 会议 follow up
by Zhangfei Gao 23 Jan '25

23 Jan '25

1. longfang 问的ctx的问题 adapter 方案 ctx 是在worker 内部的，每个worker有pool, ctx 等资源，完全解耦/不相关的比如sve 是支持异步的， hw+sve 异步是完全没问题的，代码 wd_cipher_init2_ for (i = 0; i < adapter->workers_nb; i++) { // 依次初始化各worker wd_alg_attrs_init(worker, &wd_cipher_init_attrs); switch (driver_type) { case UADK_ALG_SVE_INSTR: 分配sve的ctxs // sve, ce 其实可以合并 case UADK_ALG_HW: 分配hw的ctx wd_cipher_common_init -> wd_init_async_request_pool(&worker->pool, // 分配worker自己的pool } 另外，adapter方案的调度是选择worker, 是以worker为调度单元比如小包 sw，大包 hw，我们之前的调度，是期望每次过来能选择不同的ctx，实际中多线程需要加锁，性能跟不上，所以被废弃掉了。直接改成了alloc_sess的时候就绑定ctx，即每个线程就指定一个ctx 我理解这个其实就是ctx_allocator adapter 方案是把ctx 封装到worker内部了，即只管怎么选worker, 如果是硬件，还是按现在这样，绑定ctx的目前get_worker 选择worker 是放在算法层 2. To longfang 之前想到的问题, 在群里也提过前提是如果用户初始化的时候init(mix) 指定了选择异构调度， a, 这个代码不好改，比如已经上传了dpdk/ceph社区 b. 即使能改，init 很费时间，uadk_engine 基本是初始化的时候就init 了，即 init 一旦选择了异构调度方案就不会随便动了。 1）流模式下，不能运行一半发生切换，如何避免即调度器如何知道现在是不是流模式，算法层的信息如何传递给调度层。 ps. adapter 方案是在算法层做的，所以直接就能判断是否是流模式 2. 如何实现threshold, 很多场景就只是小包，无论怎么优化，就是指令集快调度器如何知道包的大小，需要传给调度层 ps. adapter 方案是在算法层，直接获取包的大小，这个数据结构各算法还不一样比如req->src_len vs req->in_bytes 3. 硬件出错怎么处理，fb 异步能否实现，fb方案其实我们没验证过。代码 cipher.c wd_do_cipher_async msg_id = wd_get_msg_from_pool(&wd_cipher_setting.pool, idx, (void **)&msg); //ctx_id a drv: temp_msg = wd_cipher_get_msg(qp->q_info.idx, tag); // ctx_id_a 如果fb-> drv_b temp_msg = wd_cipher_get_msg(ctx_id_b, tag); // 这里我理解拿不到msg, 可能有坑 ps：adapter方案的做法是设标记，上层retry进来选worker[1] 4. 多个加速器，如何选, hw/ce/sve 龙芳群里的回答：指定业务运行方式，配置设备驱动不懂目前的做法删掉ce的库，然后指定sve ps：adapter方案可以在uadk.conf 配置，通过命令行也可以即生成个配置文件 /tmp/pid.conf， setenv 5. 不同平台加速器不一样，920 hw, 920b hw/ce/sve, 920c? 一套模式/算法在一个平台上调好，其他平台是否适用龙芳群里的回答：自动适配，不区分硬件平台不同算法可能有区别比如rsa 就是硬件好 cipher/digest 就是小包指令集更好。 6. 未来如果有办法读取带宽，能否/怎么支持。这个龙芳群里回答过带宽查询功能具备以后，通过驱动的get_usage()直接读取，然后再新建一个调度器根据设备利用率实施调度 ps：adapter方案可能做法是get_worker里面加if/else 【腾讯文档】adapter方案 https://docs.qq.com/doc/DRUxFaVJDdHJFUWdM

2 3

2025/1/22 AccLIb SIG 双周例会 11:00 - 12:00
by openEuler conference 21 Jan '25

21 Jan '25

您好！ sig-AccLib 邀请您参加 2025-01-22 11:00 召开的Zoom会议会议主题：2025/1/22 AccLIb SIG 双周例会 11:00 - 12:00 会议内容：会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet 会议链接：https://us06web.zoom.us/j/89685603297?pwd=oTjA3Nd99NBqEQbPVD7aYubeMganRz.1 会议纪要：https://etherpad.openeuler.org/p/sig-AccLib-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! sig-AccLib invites you to attend the Zoom conference will be held at 2025-01-22 11:00, The subject of the conference is 2025/1/22 AccLIb SIG 双周例会 11:00 - 12:00, Summary: 会议链接：linaro-org.zoom.us/j/91879279131 会议纪要：etherpad.openeuler.org/p/sig-AccLib-meet You can join the meeting at https://us06web.zoom.us/j/89685603297?pwd=oTjA3Nd99NBqEQbPVD7aYubeMganRz.1. Add topics at https://etherpad.openeuler.org/p/sig-AccLib-meetings. More information: https://www.openeuler.org/en/

1 0