From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga01.intel.com (mga01.intel.com [192.55.52.88]) by dpdk.org (Postfix) with ESMTP id 499B22BA3 for ; Fri, 4 May 2018 17:56:10 +0200 (CEST) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga005.jf.intel.com ([10.7.209.41]) by fmsmga101.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 04 May 2018 08:56:08 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.49,363,1520924400"; d="scan'208";a="221659230" Received: from irsmsx106.ger.corp.intel.com ([163.33.3.31]) by orsmga005.jf.intel.com with ESMTP; 04 May 2018 08:56:06 -0700 Received: from irsmsx105.ger.corp.intel.com ([169.254.7.214]) by IRSMSX106.ger.corp.intel.com ([169.254.8.125]) with mapi id 14.03.0319.002; Fri, 4 May 2018 16:56:05 +0100 From: "Ananyev, Konstantin" To: "Guo, Jia" , "stephen@networkplumber.org" , "Richardson, Bruce" , "Yigit, Ferruh" , "gaetan.rivet@6wind.com" , "Wu, Jingjing" , "thomas@monjalon.net" , "motih@mellanox.com" , "matan@mellanox.com" , "Van Haaren, Harry" , "Tan, Jianfeng" CC: "jblunck@infradead.org" , "shreyansh.jain@nxp.com" , "dev@dpdk.org" , "Zhang, Helin" Thread-Topic: [PATCH V21 2/4] eal: add failure handle mechanism for hot plug Thread-Index: AQHT4syJCEBEclxmNk6aelBVPgf1/aQfukwA Date: Fri, 4 May 2018 15:56:04 +0000 Message-ID: <2601191342CEEE43887BDE71AB977258AEDC28C9@irsmsx105.ger.corp.intel.com> References: <1498711073-42917-1-git-send-email-jia.guo@intel.com> <1525344524-26946-1-git-send-email-jia.guo@intel.com> <1525344524-26946-3-git-send-email-jia.guo@intel.com> In-Reply-To: <1525344524-26946-3-git-send-email-jia.guo@intel.com> Accept-Language: en-IE, en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-titus-metadata-40: eyJDYXRlZ29yeUxhYmVscyI6IiIsIk1ldGFkYXRhIjp7Im5zIjoiaHR0cDpcL1wvd3d3LnRpdHVzLmNvbVwvbnNcL0ludGVsMyIsImlkIjoiZjM1MzYzM2YtMzQwNS00ZDU4LWFkMDEtOGQ5ZDliZTEzN2U0IiwicHJvcHMiOlt7Im4iOiJDVFBDbGFzc2lmaWNhdGlvbiIsInZhbHMiOlt7InZhbHVlIjoiQ1RQX05UIn1dfV19LCJTdWJqZWN0TGFiZWxzIjpbXSwiVE1DVmVyc2lvbiI6IjE2LjUuOS4zIiwiVHJ1c3RlZExhYmVsSGFzaCI6InhBMkdwYmRMRVBZNmpNWG1RMXVFRERwUmY2ZFNRQnR2WjVKMUVOMzR2d1k9In0= x-ctpclassification: CTP_NT dlp-product: dlpe-windows dlp-version: 11.0.200.100 dlp-reaction: no-action x-originating-ip: [163.33.239.181] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Subject: Re: [dpdk-dev] [PATCH V21 2/4] eal: add failure handle mechanism for hot plug X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 04 May 2018 15:56:10 -0000 Hi Jeff, >=20 > This patch introduces a failure handler mechanism to handle device > hot unplug event. When device be hot plug out, the device resource > become invalid, if this resource is still be unexpected read/write, > system will crash. This patch let eal help application to handle > this fault, when sigbus error occur, check the failure address and > accordingly remap the invalid memory for the corresponding device, > that could guaranty the application not to be shut down when hot plug. >=20 > Signed-off-by: Jeff Guo > --- > v21->v20: > sync failure hanlde to fix multiple process issue > --- > lib/librte_eal/linuxapp/eal/eal_dev.c | 154 ++++++++++++++++++++++++++++= +++++- > 1 file changed, 153 insertions(+), 1 deletion(-) >=20 > diff --git a/lib/librte_eal/linuxapp/eal/eal_dev.c b/lib/librte_eal/linux= app/eal/eal_dev.c > index 1cf6aeb..3067f39 100644 > --- a/lib/librte_eal/linuxapp/eal/eal_dev.c > +++ b/lib/librte_eal/linuxapp/eal/eal_dev.c > @@ -4,6 +4,8 @@ >=20 > #include > #include > +#include > +#include > #include > #include >=20 > @@ -14,15 +16,27 @@ > #include > #include > #include > +#include > +#include > +#include >=20 > #include "eal_private.h" >=20 > static struct rte_intr_handle intr_handle =3D {.fd =3D -1 }; > static bool monitor_started; >=20 > +extern struct rte_bus_list rte_bus_list; > + > #define EAL_UEV_MSG_LEN 4096 > #define EAL_UEV_MSG_ELEM_LEN 128 >=20 > +/* spinlock for device failure process */ > +static rte_spinlock_t dev_failure_lock =3D RTE_SPINLOCK_INITIALIZER; > + > +static struct sigaction sigbus_action_old; > + > +static int sigbus_need_recover; > + > static void dev_uev_handler(__rte_unused void *param); >=20 > /* identify the system layer which reports this event. */ > @@ -34,6 +48,93 @@ enum eal_dev_event_subsystem { > }; >=20 > static int > +dev_uev_failure_process(struct rte_device *dev, void *dev_addr) > +{ > + struct rte_bus *bus; > + int ret =3D 0; > + > + if (!dev && !dev_addr) { > + return -EINVAL; > + } else if (dev) { > + bus =3D rte_bus_find_by_device_name(dev->name); > + if (bus->handle_hot_unplug) { > + /** > + * call bus ops to handle hot unplug. > + */ > + ret =3D bus->handle_hot_unplug(dev, dev_addr); > + if (ret) { > + RTE_LOG(ERR, EAL, > + "Cannot handle hot unplug " > + "for device %s " > + "on the bus %s.\n ", > + dev->name, bus->name); > + } > + } else { > + RTE_LOG(ERR, EAL, > + "Not support handle hot unplug for bus %s!\n", > + bus->name); > + ret =3D -ENOTSUP; > + } > + } else { > + TAILQ_FOREACH(bus, &rte_bus_list, next) { > + if (bus->handle_hot_unplug) { > + /** > + * call bus ops to handle hot unplug. > + */ > + ret =3D bus->handle_hot_unplug(dev, dev_addr); > + if (ret) > + RTE_LOG(ERR, EAL, > + "Cannot handle hot unplug " > + "for the device " > + "on the bus %s!\n", bus->name); > + else > + break; > + } else { > + RTE_LOG(ERR, EAL, > + "Not support handle hot unplug " > + "for bus %s!\n", bus->name); > + ret =3D -ENOTSUP; > + } > + } > + } > + return ret; > +} > + > +static void > +sigbus_action_recover(void) > +{ > + if (sigbus_need_recover) { > + sigaction(SIGBUS, &sigbus_action_old, NULL); > + sigbus_need_recover =3D 0; > + } > +} > + > +static void sigbus_handler(int signum __rte_unused, siginfo_t *info, > + void *ctx __rte_unused) > +{ > + int ret; > + > + RTE_LOG(DEBUG, EAL, "Thread[%d] catch SIGBUS, fault address:%p\n", > + (int)pthread_self(), info->si_addr); > + rte_spinlock_lock(&dev_failure_lock); > + ret =3D dev_uev_failure_process(NULL, info->si_addr); > + rte_spinlock_unlock(&dev_failure_lock); > + if (!ret) > + RTE_LOG(DEBUG, EAL, > + "Success to handle SIGBUS error for hot unplug!\n"); > + else > + rte_exit(EXIT_FAILURE, "exit for SIGBUS error!"); I still think we have to distinguish here 2 cases: 1) failure addr is not belong to any dpdk devices 2) failure addr does belong to dpdk device, but we fail to remap it. For 1) we probably need to call previous sigbus handler. For 2) we probably can only do exit(). > +} > + > +static int cmp_dev_name(const struct rte_device *dev, > + const void *_name) > +{ > + const char *name =3D _name; > + > + return strcmp(dev->name, name); > +} > + > +static int > dev_uev_socket_fd_create(void) > { > struct sockaddr_nl addr; > @@ -147,6 +248,9 @@ dev_uev_handler(__rte_unused void *param) > struct rte_dev_event uevent; > int ret; > char buf[EAL_UEV_MSG_LEN]; > + struct rte_bus *bus; > + struct rte_device *dev; > + const char *busname; >=20 > memset(&uevent, 0, sizeof(struct rte_dev_event)); > memset(buf, 0, EAL_UEV_MSG_LEN); > @@ -171,13 +275,50 @@ dev_uev_handler(__rte_unused void *param) > RTE_LOG(DEBUG, EAL, "receive uevent(name:%s, type:%d, subsystem:%d)\n", > uevent.devname, uevent.type, uevent.subsystem); >=20 > - if (uevent.devname) > + switch (uevent.subsystem) { > + case EAL_DEV_EVENT_SUBSYSTEM_PCI: > + case EAL_DEV_EVENT_SUBSYSTEM_UIO: > + busname =3D "pci"; > + break; > + default: > + break; > + } > + > + if (uevent.devname) { > + if (uevent.type =3D=3D RTE_DEV_EVENT_REMOVE) { > + bus =3D rte_bus_find_by_name(busname); > + if (bus =3D=3D NULL) { > + RTE_LOG(ERR, EAL, "Cannot find bus (%s)\n", > + uevent.devname); > + return; > + } > + dev =3D bus->find_device(NULL, cmp_dev_name, > + uevent.devname); > + if (dev =3D=3D NULL) { > + RTE_LOG(ERR, EAL, > + "Cannot find unplugged device (%s)\n", > + uevent.devname); > + return; > + } > + rte_spinlock_lock(&dev_failure_lock); > + ret =3D dev_uev_failure_process(dev, NULL); > + rte_spinlock_unlock(&dev_failure_lock); That's interrupt thread, right? I wonder could it happen that user will call device_detach() at the same mo= ment? Konstantin > + if (ret) { > + RTE_LOG(ERR, EAL, "Driver cannot remap the " > + "device (%s)\n", > + dev->name); > + return; > + } > + } > dev_callback_process(uevent.devname, uevent.type); > + } > } >=20 > int __rte_experimental > rte_dev_event_monitor_start(void) > { > + sigset_t mask; > + struct sigaction action; > int ret; >=20 > if (monitor_started) > @@ -197,6 +338,14 @@ rte_dev_event_monitor_start(void) > return -1; > } >=20 > + /* register sigbus handler */ > + sigemptyset(&mask); > + sigaddset(&mask, SIGBUS); > + action.sa_flags =3D SA_SIGINFO; > + action.sa_mask =3D mask; > + action.sa_sigaction =3D sigbus_handler; > + sigbus_need_recover =3D !sigaction(SIGBUS, &action, &sigbus_action_old)= ; > + > monitor_started =3D true; >=20 > return 0; > @@ -217,8 +366,11 @@ rte_dev_event_monitor_stop(void) > return ret; > } >=20 > + sigbus_action_recover(); > + > close(intr_handle.fd); > intr_handle.fd =3D -1; > monitor_started =3D false; > + > return 0; > } > -- > 2.7.4