From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 9083F6FA7 for ; Sun, 17 Sep 2023 20:02:24 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id EE410C433C9; Sun, 17 Sep 2023 20:02:23 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=linuxfoundation.org; s=korg; t=1694980944; bh=Tf2zcGLX2R22IhkRfMgby+zBvicJNxjCcSW3sY21fpc=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=T8aQVm7gsxc/ZGsCGXaMC8D5ViTUmvRBCHn5J0RH7oU6YAiUhweiSHT+5j1wY9fxs 8+Hxfj0bBB41hsJ3ErXjEY3F6u6CfXpjolVCUX/epNe03F0lDzQlesKGEHnxDr5Sj0 Q18HhTGvjNY8dFVEYFePPdXPv1YB/bLYkBHLEkY0= From: Greg Kroah-Hartman To: stable@vger.kernel.org Cc: Greg Kroah-Hartman , patches@lists.linux.dev, Quinn Tran , Nilesh Javali , Himanshu Madhani , "Martin K. Petersen" Subject: [PATCH 6.1 024/219] scsi: qla2xxx: Fix firmware resource tracking Date: Sun, 17 Sep 2023 21:12:31 +0200 Message-ID: <20230917191041.869370875@linuxfoundation.org> X-Mailer: git-send-email 2.42.0 In-Reply-To: <20230917191040.964416434@linuxfoundation.org> References: <20230917191040.964416434@linuxfoundation.org> User-Agent: quilt/0.67 X-stable: review X-Patchwork-Hint: ignore Precedence: bulk X-Mailing-List: patches@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit 6.1-stable review patch. If anyone has any objections, please let me know. ------------------ From: Quinn Tran commit e370b64c7db96384a0886a09a9d80406e4c663d7 upstream. The storage was not draining I/Os and the work load was not spread out across different CPUs evenly. This led to firmware resource counters getting overrun on the busy CPU. This overrun prevented error recovery from happening in a timely manner. By switching the counter to atomic, it allows the count to be little more accurate to prevent the overrun. Cc: stable@vger.kernel.org Fixes: da7c21b72aa8 ("scsi: qla2xxx: Fix command flush during TMF") Signed-off-by: Quinn Tran Signed-off-by: Nilesh Javali Link: https://lore.kernel.org/r/20230821130045.34850-4-njavali@marvell.com Reviewed-by: Himanshu Madhani Signed-off-by: Martin K. Petersen Signed-off-by: Greg Kroah-Hartman --- drivers/scsi/qla2xxx/qla_def.h | 11 +++++++ drivers/scsi/qla2xxx/qla_dfs.c | 10 ++++++ drivers/scsi/qla2xxx/qla_init.c | 8 +++++ drivers/scsi/qla2xxx/qla_inline.h | 57 +++++++++++++++++++++++++++++++++++++- drivers/scsi/qla2xxx/qla_os.c | 5 ++- 5 files changed, 88 insertions(+), 3 deletions(-) --- a/drivers/scsi/qla2xxx/qla_def.h +++ b/drivers/scsi/qla2xxx/qla_def.h @@ -3740,6 +3740,16 @@ struct qla_fw_resources { u16 pad; }; +struct qla_fw_res { + u16 iocb_total; + u16 iocb_limit; + atomic_t iocb_used; + + u16 exch_total; + u16 exch_limit; + atomic_t exch_used; +}; + #define QLA_IOCB_PCT_LIMIT 95 /*Queue pair data structure */ @@ -4782,6 +4792,7 @@ struct qla_hw_data { spinlock_t sadb_lock; /* protects list */ struct els_reject elsrej; u8 edif_post_stop_cnt_down; + struct qla_fw_res fwres ____cacheline_aligned; }; #define RX_ELS_SIZE (roundup(sizeof(struct enode) + ELS_MAX_PAYLOAD, SMP_CACHE_BYTES)) --- a/drivers/scsi/qla2xxx/qla_dfs.c +++ b/drivers/scsi/qla2xxx/qla_dfs.c @@ -276,6 +276,16 @@ qla_dfs_fw_resource_cnt_show(struct seq_ seq_printf(s, "estimate exchange used[%d] high water limit [%d] n", exch_used, ha->base_qpair->fwres.exch_limit); + + if (ql2xenforce_iocb_limit == 2) { + iocbs_used = atomic_read(&ha->fwres.iocb_used); + exch_used = atomic_read(&ha->fwres.exch_used); + seq_printf(s, " estimate iocb2 used [%d] high water limit [%d]\n", + iocbs_used, ha->fwres.iocb_limit); + + seq_printf(s, " estimate exchange2 used[%d] high water limit [%d] \n", + exch_used, ha->fwres.exch_limit); + } } return 0; --- a/drivers/scsi/qla2xxx/qla_init.c +++ b/drivers/scsi/qla2xxx/qla_init.c @@ -4217,6 +4217,14 @@ void qla_init_iocb_limit(scsi_qla_host_t ha->queue_pair_map[i]->fwres.exch_used = 0; } } + + ha->fwres.iocb_total = ha->orig_fw_iocb_count; + ha->fwres.iocb_limit = (ha->orig_fw_iocb_count * QLA_IOCB_PCT_LIMIT) / 100; + ha->fwres.exch_total = ha->orig_fw_xcb_count; + ha->fwres.exch_limit = (ha->orig_fw_xcb_count * QLA_IOCB_PCT_LIMIT) / 100; + + atomic_set(&ha->fwres.iocb_used, 0); + atomic_set(&ha->fwres.exch_used, 0); } void qla_adjust_iocb_limit(scsi_qla_host_t *vha) --- a/drivers/scsi/qla2xxx/qla_inline.h +++ b/drivers/scsi/qla2xxx/qla_inline.h @@ -386,6 +386,7 @@ enum { RESOURCE_IOCB = BIT_0, RESOURCE_EXCH = BIT_1, /* exchange */ RESOURCE_FORCE = BIT_2, + RESOURCE_HA = BIT_3, }; static inline int @@ -393,7 +394,7 @@ qla_get_fw_resources(struct qla_qpair *q { u16 iocbs_used, i; u16 exch_used; - struct qla_hw_data *ha = qp->vha->hw; + struct qla_hw_data *ha = qp->hw; if (!ql2xenforce_iocb_limit) { iores->res_type = RESOURCE_NONE; @@ -428,15 +429,69 @@ qla_get_fw_resources(struct qla_qpair *q return -ENOSPC; } } + + if (ql2xenforce_iocb_limit == 2) { + if ((iores->iocb_cnt + atomic_read(&ha->fwres.iocb_used)) >= + ha->fwres.iocb_limit) { + iores->res_type = RESOURCE_NONE; + return -ENOSPC; + } + + if (iores->res_type & RESOURCE_EXCH) { + if ((iores->exch_cnt + atomic_read(&ha->fwres.exch_used)) >= + ha->fwres.exch_limit) { + iores->res_type = RESOURCE_NONE; + return -ENOSPC; + } + } + } + force: qp->fwres.iocbs_used += iores->iocb_cnt; qp->fwres.exch_used += iores->exch_cnt; + if (ql2xenforce_iocb_limit == 2) { + atomic_add(iores->iocb_cnt, &ha->fwres.iocb_used); + atomic_add(iores->exch_cnt, &ha->fwres.exch_used); + iores->res_type |= RESOURCE_HA; + } return 0; } +/* + * decrement to zero. This routine will not decrement below zero + * @v: pointer of type atomic_t + * @amount: amount to decrement from v + */ +static void qla_atomic_dtz(atomic_t *v, int amount) +{ + int c, old, dec; + + c = atomic_read(v); + for (;;) { + dec = c - amount; + if (unlikely(dec < 0)) + dec = 0; + + old = atomic_cmpxchg((v), c, dec); + if (likely(old == c)) + break; + c = old; + } +} + static inline void qla_put_fw_resources(struct qla_qpair *qp, struct iocb_resource *iores) { + struct qla_hw_data *ha = qp->hw; + + if (iores->res_type & RESOURCE_HA) { + if (iores->res_type & RESOURCE_IOCB) + qla_atomic_dtz(&ha->fwres.iocb_used, iores->iocb_cnt); + + if (iores->res_type & RESOURCE_EXCH) + qla_atomic_dtz(&ha->fwres.exch_used, iores->exch_cnt); + } + if (iores->res_type & RESOURCE_IOCB) { if (qp->fwres.iocbs_used >= iores->iocb_cnt) { qp->fwres.iocbs_used -= iores->iocb_cnt; --- a/drivers/scsi/qla2xxx/qla_os.c +++ b/drivers/scsi/qla2xxx/qla_os.c @@ -44,10 +44,11 @@ module_param(ql2xfulldump_on_mpifail, in MODULE_PARM_DESC(ql2xfulldump_on_mpifail, "Set this to take full dump on MPI hang."); -int ql2xenforce_iocb_limit = 1; +int ql2xenforce_iocb_limit = 2; module_param(ql2xenforce_iocb_limit, int, S_IRUGO | S_IWUSR); MODULE_PARM_DESC(ql2xenforce_iocb_limit, - "Enforce IOCB throttling, to avoid FW congestion. (default: 1)"); + "Enforce IOCB throttling, to avoid FW congestion. (default: 2) " + "1: track usage per queue, 2: track usage per adapter"); /* * CT6 CTX allocation cache