| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Fri, 1 Jul 2022 09:52:39 +0800
From: "yekai(A)" <yekai13@...wei.com>
To: Zhengchao Shao <shaozhengchao@...wei.com>,
<linux-crypto@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
<liulongfang@...wei.com>, <herbert@...dor.apana.org.au>,
<davem@...emloft.net>
CC: <weiyongjun1@...wei.com>, <yuehaibing@...wei.com>
Subject: Re: [PATCH v2] crypto: hisilicon/sec - don't sleep when in softirq
On 2022/6/30 20:26, Zhengchao Shao wrote:
> When kunpeng920 encryption driver is used to deencrypt and decrypt
> packets during the softirq, it is not allowed to use mutex lock. The
> kernel will report the following error:
>
> BUG: scheduling while atomic: swapper/57/0/0x00000300
> Call trace:
> dump_backtrace+0x0/0x1e4
> show_stack+0x20/0x2c
> dump_stack+0xd8/0x140
> __schedule_bug+0x68/0x80
> __schedule+0x728/0x840
> schedule+0x50/0xe0
> schedule_preempt_disabled+0x18/0x24
> __mutex_lock.constprop.0+0x594/0x5dc
> __mutex_lock_slowpath+0x1c/0x30
> mutex_lock+0x50/0x60
> sec_request_init+0x8c/0x1a0 [hisi_sec2]
> sec_process+0x28/0x1ac [hisi_sec2]
> sec_skcipher_crypto+0xf4/0x1d4 [hisi_sec2]
> sec_skcipher_encrypt+0x1c/0x30 [hisi_sec2]
> crypto_skcipher_encrypt+0x2c/0x40
> crypto_authenc_encrypt+0xc8/0xfc [authenc]
> crypto_aead_encrypt+0x2c/0x40
> echainiv_encrypt+0x144/0x1a0 [echainiv]
> crypto_aead_encrypt+0x2c/0x40
> esp_output_tail+0x348/0x5c0 [esp4]
> esp_output+0x120/0x19c [esp4]
> xfrm_output_one+0x25c/0x4d4
> xfrm_output_resume+0x6c/0x1fc
> xfrm_output+0xac/0x3c0
> xfrm4_output+0x64/0x130
> ip_build_and_send_pkt+0x158/0x20c
> tcp_v4_send_synack+0xdc/0x1f0
> tcp_conn_request+0x7d0/0x994
> tcp_v4_conn_request+0x58/0x6c
> tcp_v6_conn_request+0xf0/0x100
> tcp_rcv_state_process+0x1cc/0xd60
> tcp_v4_do_rcv+0x10c/0x250
> tcp_v4_rcv+0xfc4/0x10a4
> ip_protocol_deliver_rcu+0xf4/0x200
> ip_local_deliver_finish+0x58/0x70
> ip_local_deliver+0x68/0x120
> ip_sublist_rcv_finish+0x70/0x94
> ip_list_rcv_finish.constprop.0+0x17c/0x1d0
> ip_sublist_rcv+0x40/0xb0
> ip_list_rcv+0x140/0x1dc
> __netif_receive_skb_list_core+0x154/0x28c
> __netif_receive_skb_list+0x120/0x1a0
> netif_receive_skb_list_internal+0xe4/0x1f0
> napi_complete_done+0x70/0x1f0
> gro_cell_poll+0x9c/0xb0
> napi_poll+0xcc/0x264
> net_rx_action+0xd4/0x21c
> __do_softirq+0x130/0x358
> irq_exit+0x11c/0x13c
> __handle_domain_irq+0x88/0xf0
> gic_handle_irq+0x78/0x2c0
> el1_irq+0xb8/0x140
> arch_cpu_idle+0x18/0x40
> default_idle_call+0x5c/0x1c0
> cpuidle_idle_call+0x174/0x1b0
> do_idle+0xc8/0x160
> cpu_startup_entry+0x30/0x11c
> secondary_start_kernel+0x158/0x1e4
> softirq: huh, entered softirq 3 NET_RX 0000000093774ee4 with
> preempt_count 00000100, exited with fffffe00?
>
> V1: use spin_lock will cause soft lockup
>
> Signed-off-by: Zhengchao Shao <shaozhengchao@...wei.com>
Reviewed-by: Signed-off-by: Kai Ye <yekai13@...wei.com>
> ---
> drivers/crypto/hisilicon/sec2/sec.h | 2 +-
> drivers/crypto/hisilicon/sec2/sec_crypto.c | 20 ++++++++++----------
> 2 files changed, 11 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/crypto/hisilicon/sec2/sec.h b/drivers/crypto/hisilicon/sec2/sec.h
> index 42bb486f3b6d..d2a0bc93e752 100644
> --- a/drivers/crypto/hisilicon/sec2/sec.h
> +++ b/drivers/crypto/hisilicon/sec2/sec.h
> @@ -119,7 +119,7 @@ struct sec_qp_ctx {
> struct idr req_idr;
> struct sec_alg_res res[QM_Q_DEPTH];
> struct sec_ctx *ctx;
> - struct mutex req_lock;
> + spinlock_t req_lock;
> struct list_head backlog;
> struct hisi_acc_sgl_pool *c_in_pool;
> struct hisi_acc_sgl_pool *c_out_pool;
> diff --git a/drivers/crypto/hisilicon/sec2/sec_crypto.c b/drivers/crypto/hisilicon/sec2/sec_crypto.c
> index 6eebe739893c..71dfa7db6394 100644
> --- a/drivers/crypto/hisilicon/sec2/sec_crypto.c
> +++ b/drivers/crypto/hisilicon/sec2/sec_crypto.c
> @@ -127,11 +127,11 @@ static int sec_alloc_req_id(struct sec_req *req, struct sec_qp_ctx *qp_ctx)
> {
> int req_id;
>
> - mutex_lock(&qp_ctx->req_lock);
> + spin_lock_bh(&qp_ctx->req_lock);
>
> req_id = idr_alloc_cyclic(&qp_ctx->req_idr, NULL,
> 0, QM_Q_DEPTH, GFP_ATOMIC);
> - mutex_unlock(&qp_ctx->req_lock);
> + spin_unlock_bh(&qp_ctx->req_lock);
> if (unlikely(req_id < 0)) {
> dev_err(req->ctx->dev, "alloc req id fail!\n");
> return req_id;
> @@ -156,9 +156,9 @@ static void sec_free_req_id(struct sec_req *req)
> qp_ctx->req_list[req_id] = NULL;
> req->qp_ctx = NULL;
>
> - mutex_lock(&qp_ctx->req_lock);
> + spin_lock_bh(&qp_ctx->req_lock);
> idr_remove(&qp_ctx->req_idr, req_id);
> - mutex_unlock(&qp_ctx->req_lock);
> + spin_unlock_bh(&qp_ctx->req_lock);
> }
>
> static u8 pre_parse_finished_bd(struct bd_status *status, void *resp)
> @@ -273,7 +273,7 @@ static int sec_bd_send(struct sec_ctx *ctx, struct sec_req *req)
> !(req->flag & CRYPTO_TFM_REQ_MAY_BACKLOG))
> return -EBUSY;
>
> - mutex_lock(&qp_ctx->req_lock);
> + spin_lock_bh(&qp_ctx->req_lock);
> ret = hisi_qp_send(qp_ctx->qp, &req->sec_sqe);
>
> if (ctx->fake_req_limit <=
> @@ -281,10 +281,10 @@ static int sec_bd_send(struct sec_ctx *ctx, struct sec_req *req)
> list_add_tail(&req->backlog_head, &qp_ctx->backlog);
> atomic64_inc(&ctx->sec->debug.dfx.send_cnt);
> atomic64_inc(&ctx->sec->debug.dfx.send_busy_cnt);
> - mutex_unlock(&qp_ctx->req_lock);
> + spin_unlock_bh(&qp_ctx->req_lock);
> return -EBUSY;
> }
> - mutex_unlock(&qp_ctx->req_lock);
> + spin_unlock_bh(&qp_ctx->req_lock);
>
> if (unlikely(ret == -EBUSY))
> return -ENOBUFS;
> @@ -487,7 +487,7 @@ static int sec_create_qp_ctx(struct hisi_qm *qm, struct sec_ctx *ctx,
>
> qp->req_cb = sec_req_cb;
>
> - mutex_init(&qp_ctx->req_lock);
> + spin_lock_init(&qp_ctx->req_lock);
> idr_init(&qp_ctx->req_idr);
> INIT_LIST_HEAD(&qp_ctx->backlog);
>
> @@ -1382,7 +1382,7 @@ static struct sec_req *sec_back_req_clear(struct sec_ctx *ctx,
> {
> struct sec_req *backlog_req = NULL;
>
> - mutex_lock(&qp_ctx->req_lock);
> + spin_lock_bh(&qp_ctx->req_lock);
> if (ctx->fake_req_limit >=
> atomic_read(&qp_ctx->qp->qp_status.used) &&
> !list_empty(&qp_ctx->backlog)) {
> @@ -1390,7 +1390,7 @@ static struct sec_req *sec_back_req_clear(struct sec_ctx *ctx,
> typeof(*backlog_req), backlog_head);
> list_del(&backlog_req->backlog_head);
> }
> - mutex_unlock(&qp_ctx->req_lock);
> + spin_unlock_bh(&qp_ctx->req_lock);
>
> return backlog_req;
> }
>
Powered by blists - more mailing lists