Message ID | 1524123964-21347-4-git-send-email-xiubli@redhat.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On 04/19/2018 02:46 AM, xiubli@redhat.com wrote: > @@ -1572,13 +1579,16 @@ static int tcmu_wait_genl_cmd_reply(struct tcmu_dev *udev) > if (udev->nl_reply_supported <= 0) > return 0; > > + spin_lock(&udev->nl_cmd_lock); > + nl_cmd->waiter++; I think this will allow races. 1. It is not likely, but if userspace crashed after the nl event was sent to userspace (maybe it crashed during the handling of the event), and then restarted before we did waiter++ we would miss this cmd and not clean it up. 2. With your userspace patch, we reset outstanding events with watier incremented, but the event could still be queued in the netlink socket. When runner starts to read what is in the socket it will start to handle events that had waiter incremented but the reset operation has force failed. I think we could add a block/unblock type of operation where we do 1. block. Fail new cmds or put the reqeuster to sleep. 2. userspace dequeues events in nl socket and handles them. 3. userspace resets nl like we do like in this patch and fail outstanding cmds that were already dequeued and we have no info about. 4. unblock. Or maybe I think we could add a list/queue and we could figure out what has been dequeued vs run vs is waiting for a completion. > + spin_unlock(&udev->nl_cmd_lock); > + > pr_debug("sleeping for nl reply\n"); > - wait_for_completion(&nl_cmd->complete); > + wait_event(udev->complete_wq, nl_cmd->status != 1); > > spin_lock(&udev->nl_cmd_lock); > nl_cmd->cmd = TCMU_CMD_UNSPEC; > ret = nl_cmd->status; > - nl_cmd->status = 0; > spin_unlock(&udev->nl_cmd_lock); > > wake_up_all(&udev->nl_cmd_wq); > @@ -2366,6 +2376,54 @@ static ssize_t tcmu_reset_ring_store(struct config_item *item, const char *page, > NULL, > }; > > +static int tcmu_complete_wake_up_iter(struct se_device *se_dev, void *data) > +{ I think it's best to do this after we know it's a tcmu device > + struct tcmu_dev *udev = TCMU_DEV(se_dev); > + struct tcmu_nl_cmd *nl_cmd; > + > + if (se_dev->transport != &tcmu_ops) > + return 0; > + -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 2018/5/3 3:53, Mike Christie wrote: > On 04/19/2018 02:46 AM, xiubli@redhat.com wrote: >> @@ -1572,13 +1579,16 @@ static int tcmu_wait_genl_cmd_reply(struct tcmu_dev *udev) >> if (udev->nl_reply_supported <= 0) >> return 0; >> >> + spin_lock(&udev->nl_cmd_lock); >> + nl_cmd->waiter++; > I think this will allow races. > > 1. It is not likely, but if userspace crashed after the nl event was > sent to userspace (maybe it crashed during the handling of the event), > and then restarted before we did waiter++ we would miss this cmd and not > clean it up. Yes, in theory it is. > 2. With your userspace patch, we reset outstanding events with watier > incremented, but the event could still be queued in the netlink socket. > When runner starts to read what is in the socket it will start to handle > events that had waiter incremented but the reset operation has force failed. > > I think we could add a block/unblock type of operation where we do > > 1. block. Fail new cmds or put the reqeuster to sleep. > 2. userspace dequeues events in nl socket and handles them. > 3. userspace resets nl like we do like in this patch and fail > outstanding cmds that were already dequeued and we have no info about. > 4. unblock. IMO, this is a better choice. We'd better fail the new cmds, because there maybe some commands have dependency, such as if the RECONFIG cmd follows the ADD cmd and the ADD failed after crash. Just do: 1. block. a), Fail new cmds, because the usespace hasn't been ready. b), Wake up the waiting cmds then fail them, because only when the device has an unreplied cmd it will have waiting cmds. Because we didn't know whether the old cmd is blocked in nl socket or already be lost. 2. userspace dequeues events in nl socket and handles them if there has. 3. userspace resets and fails outstanding cmds that were already dequeued and we have no info about(the lost cmds). 4. unblock. BRs > Or maybe I think we could add a list/queue and we could figure out what > has been dequeued vs run vs is waiting for a completion. > > >> + spin_unlock(&udev->nl_cmd_lock); >> + >> pr_debug("sleeping for nl reply\n"); >> - wait_for_completion(&nl_cmd->complete); >> + wait_event(udev->complete_wq, nl_cmd->status != 1); >> >> spin_lock(&udev->nl_cmd_lock); >> nl_cmd->cmd = TCMU_CMD_UNSPEC; >> ret = nl_cmd->status; >> - nl_cmd->status = 0; >> spin_unlock(&udev->nl_cmd_lock); >> >> wake_up_all(&udev->nl_cmd_wq); >> @@ -2366,6 +2376,54 @@ static ssize_t tcmu_reset_ring_store(struct config_item *item, const char *page, >> NULL, >> }; >> >> +static int tcmu_complete_wake_up_iter(struct se_device *se_dev, void *data) >> +{ > I think it's best to do this after we know it's a tcmu device > > >> + struct tcmu_dev *udev = TCMU_DEV(se_dev); >> + struct tcmu_nl_cmd *nl_cmd; >> + >> + if (se_dev->transport != &tcmu_ops) >> + return 0; >> + -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
============== INFO: task targetctl:22655 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. targetctl D ffff880169718fd0 0 22655 17249 0x00000080 Call Trace: [<ffffffff816ab6d9>] schedule+0x29/0x70 [<ffffffff816a90e9>] schedule_timeout+0x239/0x2c0 [<ffffffff81574d42>] ? skb_release_data+0xf2/0x140 [<ffffffff816aba8d>] wait_for_completion+0xfd/0x140 [<ffffffff810c6440>] ? wake_up_state+0x20/0x20 [<ffffffffc0159f5a>] tcmu_netlink_event+0x26a/0x3a0 [target_core_user] [<ffffffff810b34b0>] ? wake_up_atomic_t+0x30/0x30 [<ffffffffc015a2c6>] tcmu_configure_device+0x236/0x350 [target_core_user] [<ffffffffc05085df>] target_configure_device+0x3f/0x3b0 [target_core_mod] [<ffffffffc0502e7c>] target_core_store_dev_enable+0x2c/0x60 [target_core_mod] [<ffffffffc0501244>] target_core_dev_store+0x24/0x40 [target_core_mod] [<ffffffff8128a0e4>] configfs_write_file+0xc4/0x130 [<ffffffff81202aed>] vfs_write+0xbd/0x1e0 [<ffffffff812038ff>] SyS_write+0x7f/0xe0 [<ffffffff816b89fd>] system_call_fastpath+0x16/0x1b ============== Be careful of using this, it could reset the normal netlink requesting operations, so we should use this only when the user space daemon from starting and just before the daemon could receive and handle the nl requests. Changes since v1(suggested by Mike Christie): v2: - Makes the reset per device. v3: - Remove nl_cmd->complete, use status instead - Fix lock issue - Check if a nl command is even waiting before trying to wake up v4: - Add module wide action support and make the reset netlink for the module wide Signed-off-by: Xiubo Li <xiubli@redhat.com> --- drivers/target/target_core_user.c | 77 ++++++++++++++++++++++++++++++++++----- 1 file changed, 68 insertions(+), 9 deletions(-) diff --git a/drivers/target/target_core_user.c b/drivers/target/target_core_user.c index 4ad89ea..f520933 100644 --- a/drivers/target/target_core_user.c +++ b/drivers/target/target_core_user.c @@ -96,6 +96,7 @@ static u8 tcmu_kern_cmd_reply_supported; static struct device *tcmu_root_device; +static struct target_backend_ops tcmu_ops; struct tcmu_hba { u32 host_id; @@ -104,8 +105,7 @@ struct tcmu_hba { #define TCMU_CONFIG_LEN 256 struct tcmu_nl_cmd { - /* wake up thread waiting for reply */ - struct completion complete; + unsigned int waiter; int cmd; int status; }; @@ -159,9 +159,12 @@ struct tcmu_dev { spinlock_t nl_cmd_lock; struct tcmu_nl_cmd curr_nl_cmd; - /* wake up threads waiting on curr_nl_cmd */ + /* wake up threads waiting on nl_cmd_wq */ wait_queue_head_t nl_cmd_wq; + /* complete thread waiting complete_wq */ + wait_queue_head_t complete_wq; + char dev_config[TCMU_CONFIG_LEN]; int nl_reply_supported; @@ -307,11 +310,14 @@ static int tcmu_genl_cmd_done(struct genl_info *info, int completed_cmd) nl_cmd->status = rc; } - spin_unlock(&udev->nl_cmd_lock); if (!is_removed) target_undepend_item(&dev->dev_group.cg_item); - if (!ret) - complete(&nl_cmd->complete); + if (!ret && nl_cmd->waiter) { + nl_cmd->waiter--; + wake_up(&udev->complete_wq); + } + spin_unlock(&udev->nl_cmd_lock); + return ret; } @@ -1258,6 +1264,7 @@ static struct se_device *tcmu_alloc_device(struct se_hba *hba, const char *name) timer_setup(&udev->cmd_timer, tcmu_cmd_timedout, 0); init_waitqueue_head(&udev->nl_cmd_wq); + init_waitqueue_head(&udev->complete_wq); spin_lock_init(&udev->nl_cmd_lock); INIT_RADIX_TREE(&udev->data_blocks, GFP_KERNEL); @@ -1555,7 +1562,7 @@ static void tcmu_init_genl_cmd_reply(struct tcmu_dev *udev, int cmd) memset(nl_cmd, 0, sizeof(*nl_cmd)); nl_cmd->cmd = cmd; - init_completion(&nl_cmd->complete); + nl_cmd->status = 1; spin_unlock(&udev->nl_cmd_lock); } @@ -1572,13 +1579,16 @@ static int tcmu_wait_genl_cmd_reply(struct tcmu_dev *udev) if (udev->nl_reply_supported <= 0) return 0; + spin_lock(&udev->nl_cmd_lock); + nl_cmd->waiter++; + spin_unlock(&udev->nl_cmd_lock); + pr_debug("sleeping for nl reply\n"); - wait_for_completion(&nl_cmd->complete); + wait_event(udev->complete_wq, nl_cmd->status != 1); spin_lock(&udev->nl_cmd_lock); nl_cmd->cmd = TCMU_CMD_UNSPEC; ret = nl_cmd->status; - nl_cmd->status = 0; spin_unlock(&udev->nl_cmd_lock); wake_up_all(&udev->nl_cmd_wq); @@ -2366,6 +2376,54 @@ static ssize_t tcmu_reset_ring_store(struct config_item *item, const char *page, NULL, }; +static int tcmu_complete_wake_up_iter(struct se_device *se_dev, void *data) +{ + struct tcmu_dev *udev = TCMU_DEV(se_dev); + struct tcmu_nl_cmd *nl_cmd; + + if (se_dev->transport != &tcmu_ops) + return 0; + + spin_lock(&udev->nl_cmd_lock); + nl_cmd = &udev->curr_nl_cmd; + if (nl_cmd->waiter) { + nl_cmd->waiter--; + nl_cmd->status = -EINTR; + wake_up(&udev->complete_wq); + } + spin_unlock(&udev->nl_cmd_lock); + + return 0; +} + +static ssize_t tcmu_reset_netlink_store(struct config_item *item, const char *page, + size_t count) +{ + u8 val; + int ret; + + ret = kstrtou8(page, 0, &val); + if (ret < 0) + return ret; + + if (val != 1) { + pr_err("Invalid block value %d\n", val); + return -EINVAL; + } + + ret = target_for_each_device(tcmu_complete_wake_up_iter, NULL); + if (ret) + return ret; + + return count; +} +CONFIGFS_ATTR_WO(tcmu_, reset_netlink); + +static struct configfs_attribute *tcmu_mod_action_attrs[] = { + &tcmu_attr_reset_netlink, + NULL, +}; + static struct target_backend_ops tcmu_ops = { .name = "user", .owner = THIS_MODULE, @@ -2382,6 +2440,7 @@ static ssize_t tcmu_reset_ring_store(struct config_item *item, const char *page, .get_device_type = sbc_get_device_type, .get_blocks = tcmu_get_blocks, .tb_dev_action_attrs = tcmu_action_attrs, + .tb_mod_action_attrs = tcmu_mod_action_attrs, }; static void find_free_blocks(void)
From: Xiubo Li <xiubli@redhat.com> This patch adds 1 tcmu attr to reset and complete all the blocked netlink waiting threads. It's used when the userspace daemon like tcmu-runner has crashed or forced to shutdown just before the netlink requests be replied to the kernel, then the netlink requeting threads will get stuck forever. We must reboot the machine to recover from it and by this the rebootng is not a must then. The Call Trace will be something like: