- Acc - mailweb.openeuler.org

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

[PATCH 00/11] uadk_prov: support new algs
by Qi Tao 20 Jan '25

20 Jan '25

*** BLURB HERE *** Chenghai Huang (4): uadk_provider/aead: cleanup functions in aead uadk_provider: fix the fork function of the digest uadk_provider: fix the ctx copy function of the digest uadk_provider: code cleanup for digest and aead Weili Qian (3): uadk_provider/ecdsa: support ecdsa signature and verify uadk_async: check type and ctx in task uadk_provider: use default provider to get random data Zhiqi Song (4): uadk_engine/rsa: fix memory leak of pkey meth uadk_engine: fix memory leak in fork scene uadk_provider: support X25519 uadk_engine/rsa: add rsa finish method src/Makefile.am | 2 +- src/uadk_async.c | 5 +- src/uadk_engine_init.c | 2 +- src/uadk_prov.h | 5 + src/uadk_prov_aead.c | 92 ++- src/uadk_prov_cipher.c | 3 + src/uadk_prov_der_writer.c | 112 +++- src/uadk_prov_der_writer.h | 2 + src/uadk_prov_digest.c | 235 ++++++-- src/uadk_prov_ecdsa.c | 1151 ++++++++++++++++++++++++++++++++++++ src/uadk_prov_ecx.c | 517 ++++++++++++++-- src/uadk_prov_init.c | 17 +- src/uadk_prov_pkey.c | 28 +- src/uadk_prov_pkey.h | 3 + src/uadk_rsa.c | 11 + 15 files changed, 2003 insertions(+), 182 deletions(-) create mode 100644 src/uadk_prov_ecdsa.c -- 2.33.0

1 0