From 77f88796cee819b9c4562b0b6b44691b3b7755b1 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Thu, 16 Mar 2017 16:54:24 -0400 Subject: cgroup, kthread: close race window where new kthreads can be migrated to non-root cgroups Creation of a kthread goes through a couple interlocked stages between the kthread itself and its creator. Once the new kthread starts running, it initializes itself and wakes up the creator. The creator then can further configure the kthread and then let it start doing its job by waking it up. In this configuration-by-creator stage, the creator is the only one that can wake it up but the kthread is visible to userland. When altering the kthread's attributes from userland is allowed, this is fine; however, for cases where CPU affinity is critical, kthread_bind() is used to first disable affinity changes from userland and then set the affinity. This also prevents the kthread from being migrated into non-root cgroups as that can affect the CPU affinity and many other things. Unfortunately, the cgroup side of protection is racy. While the PF_NO_SETAFFINITY flag prevents further migrations, userland can win the race before the creator sets the flag with kthread_bind() and put the kthread in a non-root cgroup, which can lead to all sorts of problems including incorrect CPU affinity and starvation. This bug got triggered by userland which periodically tries to migrate all processes in the root cpuset cgroup to a non-root one. Per-cpu workqueue workers got caught while being created and ended up with incorrected CPU affinity breaking concurrency management and sometimes stalling workqueue execution. This patch adds task->no_cgroup_migration which disallows the task to be migrated by userland. kthreadd starts with the flag set making every child kthread start in the root cgroup with migration disallowed. The flag is cleared after the kthread finishes initialization by which time PF_NO_SETAFFINITY is set if the kthread should stay in the root cgroup. It'd be better to wait for the initialization instead of failing but I couldn't think of a way of implementing that without adding either a new PF flag, or sleeping and retrying from waiting side. Even if userland depends on changing cgroup membership of a kthread, it either has to be synchronized with kthread_create() or periodically repeat, so it's unlikely that this would break anything. v2: Switch to a simpler implementation using a new task_struct bit field suggested by Oleg. Signed-off-by: Tejun Heo Suggested-by: Oleg Nesterov Cc: Linus Torvalds Cc: Andrew Morton Cc: Peter Zijlstra (Intel) Cc: Thomas Gleixner Reported-and-debugged-by: Chris Mason Cc: stable@vger.kernel.org # v4.3+ (we can't close the race on < v4.3) Signed-off-by: Tejun Heo --- include/linux/cgroup.h | 21 +++++++++++++++++++++ include/linux/sched.h | 4 ++++ 2 files changed, 25 insertions(+) (limited to 'include') diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h index f6b43fbb141c..af9c86e958bd 100644 --- a/include/linux/cgroup.h +++ b/include/linux/cgroup.h @@ -570,6 +570,25 @@ static inline void pr_cont_cgroup_path(struct cgroup *cgrp) pr_cont_kernfs_path(cgrp->kn); } +static inline void cgroup_init_kthreadd(void) +{ + /* + * kthreadd is inherited by all kthreads, keep it in the root so + * that the new kthreads are guaranteed to stay in the root until + * initialization is finished. + */ + current->no_cgroup_migration = 1; +} + +static inline void cgroup_kthread_ready(void) +{ + /* + * This kthread finished initialization. The creator should have + * set PF_NO_SETAFFINITY if this kthread should stay in the root. + */ + current->no_cgroup_migration = 0; +} + #else /* !CONFIG_CGROUPS */ struct cgroup_subsys_state; @@ -590,6 +609,8 @@ static inline void cgroup_free(struct task_struct *p) {} static inline int cgroup_init_early(void) { return 0; } static inline int cgroup_init(void) { return 0; } +static inline void cgroup_init_kthreadd(void) {} +static inline void cgroup_kthread_ready(void) {} static inline bool task_under_cgroup_hierarchy(struct task_struct *task, struct cgroup *ancestor) diff --git a/include/linux/sched.h b/include/linux/sched.h index d67eee84fd43..4cf9a59a4d08 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -604,6 +604,10 @@ struct task_struct { #ifdef CONFIG_COMPAT_BRK unsigned brk_randomized:1; #endif +#ifdef CONFIG_CGROUPS + /* disallow userland-initiated cgroup migration */ + unsigned no_cgroup_migration:1; +#endif unsigned long atomic_flags; /* Flags requiring atomic access. */ -- cgit v1.2.3 From 5659495a7a1455665ce1466d156597ad1bda8772 Mon Sep 17 00:00:00 2001 From: Naohiro Aota Date: Fri, 24 Mar 2017 23:04:44 -0700 Subject: uapi: add missing install of userio.h While commit 5523662edd4f ("Input: add userio module") added userio.h under the uapi/ directory, it forgot to add the header file to Kbuild. Thus, the file was missing from header installation. Signed-off-by: Naohiro Aota Reviewed-by: Lyude Paul Signed-off-by: Dmitry Torokhov --- include/uapi/linux/Kbuild | 1 + 1 file changed, 1 insertion(+) (limited to 'include') diff --git a/include/uapi/linux/Kbuild b/include/uapi/linux/Kbuild index f330ba4547cf..b4a9a1891db6 100644 --- a/include/uapi/linux/Kbuild +++ b/include/uapi/linux/Kbuild @@ -437,6 +437,7 @@ header-y += unistd.h header-y += unix_diag.h header-y += usbdevice_fs.h header-y += usbip.h +header-y += userio.h header-y += utime.h header-y += utsname.h header-y += uuid.h -- cgit v1.2.3 From 49cb77e297dc611a1b795cfeb79452b3002bd331 Mon Sep 17 00:00:00 2001 From: Nicholas Bellinger Date: Mon, 27 Mar 2017 16:12:43 -0700 Subject: target: Avoid mappedlun symlink creation during lun shutdown This patch closes a race between se_lun deletion during configfs unlink in target_fabric_port_unlink() -> core_dev_del_lun() -> core_tpg_remove_lun(), when transport_clear_lun_ref() blocks waiting for percpu_ref RCU grace period to finish, but a new NodeACL mappedlun is added before the RCU grace period has completed. This can happen in target_fabric_mappedlun_link() because it only checks for se_lun->lun_se_dev, which is not cleared until after transport_clear_lun_ref() percpu_ref RCU grace period finishes. This bug originally manifested as NULL pointer dereference OOPsen in target_stat_scsi_att_intr_port_show_attr_dev() on v4.1.y code, because it dereferences lun->lun_se_dev without a explicit NULL pointer check. In post v4.1 code with target-core RCU conversion, the code in target_stat_scsi_att_intr_port_show_attr_dev() no longer uses se_lun->lun_se_dev, but the same race still exists. To address the bug, go ahead and set se_lun>lun_shutdown as early as possible in core_tpg_remove_lun(), and ensure new NodeACL mappedlun creation in target_fabric_mappedlun_link() fails during se_lun shutdown. Reported-by: James Shen Cc: James Shen Tested-by: James Shen Cc: stable@vger.kernel.org # 3.10+ Signed-off-by: Nicholas Bellinger --- drivers/target/target_core_fabric_configfs.c | 5 +++++ drivers/target/target_core_tpg.c | 4 ++++ include/target/target_core_base.h | 1 + 3 files changed, 10 insertions(+) (limited to 'include') diff --git a/drivers/target/target_core_fabric_configfs.c b/drivers/target/target_core_fabric_configfs.c index d8a16ca6baa5..d1e6cab8e3d3 100644 --- a/drivers/target/target_core_fabric_configfs.c +++ b/drivers/target/target_core_fabric_configfs.c @@ -92,6 +92,11 @@ static int target_fabric_mappedlun_link( pr_err("Source se_lun->lun_se_dev does not exist\n"); return -EINVAL; } + if (lun->lun_shutdown) { + pr_err("Unable to create mappedlun symlink because" + " lun->lun_shutdown=true\n"); + return -EINVAL; + } se_tpg = lun->lun_tpg; nacl_ci = &lun_acl_ci->ci_parent->ci_group->cg_item; diff --git a/drivers/target/target_core_tpg.c b/drivers/target/target_core_tpg.c index 6fb191914f45..dfaef4d3b2d2 100644 --- a/drivers/target/target_core_tpg.c +++ b/drivers/target/target_core_tpg.c @@ -642,6 +642,8 @@ void core_tpg_remove_lun( */ struct se_device *dev = rcu_dereference_raw(lun->lun_se_dev); + lun->lun_shutdown = true; + core_clear_lun_from_tpg(lun, tpg); /* * Wait for any active I/O references to percpu se_lun->lun_ref to @@ -663,6 +665,8 @@ void core_tpg_remove_lun( } if (!(dev->se_hba->hba_flags & HBA_FLAGS_INTERNAL_USE)) hlist_del_rcu(&lun->link); + + lun->lun_shutdown = false; mutex_unlock(&tpg->tpg_lun_mutex); percpu_ref_exit(&lun->lun_ref); diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h index 4b784b6e21c0..2e282461cfa5 100644 --- a/include/target/target_core_base.h +++ b/include/target/target_core_base.h @@ -705,6 +705,7 @@ struct se_lun { u64 unpacked_lun; #define SE_LUN_LINK_MAGIC 0xffff7771 u32 lun_link_magic; + bool lun_shutdown; bool lun_access_ro; u32 lun_index; -- cgit v1.2.3 From fa7e25cf13a6d0b82b5ed1008246f44d42e8422c Mon Sep 17 00:00:00 2001 From: Nicholas Bellinger Date: Sun, 30 Oct 2016 17:28:16 -0700 Subject: target: Fix unknown fabric callback queue-full errors This patch fixes a set of queue-full response handling bugs, where outgoing responses are leaked when a fabric driver is propagating non -EAGAIN or -ENOMEM errors to target-core. It introduces TRANSPORT_COMPLETE_QF_ERR state used to signal when CHECK_CONDITION status should be generated, when fabric driver ->write_pending(), ->queue_data_in(), or ->queue_status() callbacks fail with non -EAGAIN or -ENOMEM errors, and data-transfer should not be retried. Note all fabric driver -EAGAIN and -ENOMEM errors are still retried indefinately with associated data-transfer callbacks, following existing queue-full logic. Also fix two missing ->queue_status() queue-full cases related to CMD_T_ABORTED w/ TAS status handling. Reported-by: Potnuri Bharat Teja Reviewed-by: Potnuri Bharat Teja Tested-by: Potnuri Bharat Teja Cc: Potnuri Bharat Teja Reported-by: Steve Wise Cc: Steve Wise Cc: Sagi Grimberg Signed-off-by: Nicholas Bellinger --- drivers/target/target_core_transport.c | 102 ++++++++++++++++++++++----------- include/target/target_core_base.h | 1 + 2 files changed, 69 insertions(+), 34 deletions(-) (limited to 'include') diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c index b1a3cdb29468..a0cd56ee5fe9 100644 --- a/drivers/target/target_core_transport.c +++ b/drivers/target/target_core_transport.c @@ -64,8 +64,9 @@ struct kmem_cache *t10_alua_lba_map_cache; struct kmem_cache *t10_alua_lba_map_mem_cache; static void transport_complete_task_attr(struct se_cmd *cmd); +static int translate_sense_reason(struct se_cmd *cmd, sense_reason_t reason); static void transport_handle_queue_full(struct se_cmd *cmd, - struct se_device *dev); + struct se_device *dev, int err, bool write_pending); static int transport_put_cmd(struct se_cmd *cmd); static void target_complete_ok_work(struct work_struct *work); @@ -804,7 +805,8 @@ void target_qf_do_work(struct work_struct *work) if (cmd->t_state == TRANSPORT_COMPLETE_QF_WP) transport_write_pending_qf(cmd); - else if (cmd->t_state == TRANSPORT_COMPLETE_QF_OK) + else if (cmd->t_state == TRANSPORT_COMPLETE_QF_OK || + cmd->t_state == TRANSPORT_COMPLETE_QF_ERR) transport_complete_qf(cmd); } } @@ -1719,7 +1721,7 @@ void transport_generic_request_failure(struct se_cmd *cmd, } trace_target_cmd_complete(cmd); ret = cmd->se_tfo->queue_status(cmd); - if (ret == -EAGAIN || ret == -ENOMEM) + if (ret) goto queue_full; goto check_stop; default: @@ -1730,7 +1732,7 @@ void transport_generic_request_failure(struct se_cmd *cmd, } ret = transport_send_check_condition_and_sense(cmd, sense_reason, 0); - if (ret == -EAGAIN || ret == -ENOMEM) + if (ret) goto queue_full; check_stop: @@ -1739,8 +1741,7 @@ check_stop: return; queue_full: - cmd->t_state = TRANSPORT_COMPLETE_QF_OK; - transport_handle_queue_full(cmd, cmd->se_dev); + transport_handle_queue_full(cmd, cmd->se_dev, ret, false); } EXPORT_SYMBOL(transport_generic_request_failure); @@ -1977,13 +1978,29 @@ static void transport_complete_qf(struct se_cmd *cmd) int ret = 0; transport_complete_task_attr(cmd); + /* + * If a fabric driver ->write_pending() or ->queue_data_in() callback + * has returned neither -ENOMEM or -EAGAIN, assume it's fatal and + * the same callbacks should not be retried. Return CHECK_CONDITION + * if a scsi_status is not already set. + * + * If a fabric driver ->queue_status() has returned non zero, always + * keep retrying no matter what.. + */ + if (cmd->t_state == TRANSPORT_COMPLETE_QF_ERR) { + if (cmd->scsi_status) + goto queue_status; - if (cmd->se_cmd_flags & SCF_TRANSPORT_TASK_SENSE) { - trace_target_cmd_complete(cmd); - ret = cmd->se_tfo->queue_status(cmd); - goto out; + cmd->se_cmd_flags |= SCF_EMULATED_TASK_SENSE; + cmd->scsi_status = SAM_STAT_CHECK_CONDITION; + cmd->scsi_sense_length = TRANSPORT_SENSE_BUFFER; + translate_sense_reason(cmd, TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE); + goto queue_status; } + if (cmd->se_cmd_flags & SCF_TRANSPORT_TASK_SENSE) + goto queue_status; + switch (cmd->data_direction) { case DMA_FROM_DEVICE: if (cmd->scsi_status) @@ -2007,19 +2024,33 @@ queue_status: break; } -out: if (ret < 0) { - transport_handle_queue_full(cmd, cmd->se_dev); + transport_handle_queue_full(cmd, cmd->se_dev, ret, false); return; } transport_lun_remove_cmd(cmd); transport_cmd_check_stop_to_fabric(cmd); } -static void transport_handle_queue_full( - struct se_cmd *cmd, - struct se_device *dev) +static void transport_handle_queue_full(struct se_cmd *cmd, struct se_device *dev, + int err, bool write_pending) { + /* + * -EAGAIN or -ENOMEM signals retry of ->write_pending() and/or + * ->queue_data_in() callbacks from new process context. + * + * Otherwise for other errors, transport_complete_qf() will send + * CHECK_CONDITION via ->queue_status() instead of attempting to + * retry associated fabric driver data-transfer callbacks. + */ + if (err == -EAGAIN || err == -ENOMEM) { + cmd->t_state = (write_pending) ? TRANSPORT_COMPLETE_QF_WP : + TRANSPORT_COMPLETE_QF_OK; + } else { + pr_warn_ratelimited("Got unknown fabric queue status: %d\n", err); + cmd->t_state = TRANSPORT_COMPLETE_QF_ERR; + } + spin_lock_irq(&dev->qf_cmd_lock); list_add_tail(&cmd->se_qf_node, &cmd->se_dev->qf_cmd_list); atomic_inc_mb(&dev->dev_qf_count); @@ -2083,7 +2114,7 @@ static void target_complete_ok_work(struct work_struct *work) WARN_ON(!cmd->scsi_status); ret = transport_send_check_condition_and_sense( cmd, 0, 1); - if (ret == -EAGAIN || ret == -ENOMEM) + if (ret) goto queue_full; transport_lun_remove_cmd(cmd); @@ -2109,7 +2140,7 @@ static void target_complete_ok_work(struct work_struct *work) } else if (rc) { ret = transport_send_check_condition_and_sense(cmd, rc, 0); - if (ret == -EAGAIN || ret == -ENOMEM) + if (ret) goto queue_full; transport_lun_remove_cmd(cmd); @@ -2134,7 +2165,7 @@ queue_rsp: if (target_read_prot_action(cmd)) { ret = transport_send_check_condition_and_sense(cmd, cmd->pi_err, 0); - if (ret == -EAGAIN || ret == -ENOMEM) + if (ret) goto queue_full; transport_lun_remove_cmd(cmd); @@ -2144,7 +2175,7 @@ queue_rsp: trace_target_cmd_complete(cmd); ret = cmd->se_tfo->queue_data_in(cmd); - if (ret == -EAGAIN || ret == -ENOMEM) + if (ret) goto queue_full; break; case DMA_TO_DEVICE: @@ -2157,7 +2188,7 @@ queue_rsp: atomic_long_add(cmd->data_length, &cmd->se_lun->lun_stats.tx_data_octets); ret = cmd->se_tfo->queue_data_in(cmd); - if (ret == -EAGAIN || ret == -ENOMEM) + if (ret) goto queue_full; break; } @@ -2166,7 +2197,7 @@ queue_rsp: queue_status: trace_target_cmd_complete(cmd); ret = cmd->se_tfo->queue_status(cmd); - if (ret == -EAGAIN || ret == -ENOMEM) + if (ret) goto queue_full; break; default: @@ -2180,8 +2211,8 @@ queue_status: queue_full: pr_debug("Handling complete_ok QUEUE_FULL: se_cmd: %p," " data_direction: %d\n", cmd, cmd->data_direction); - cmd->t_state = TRANSPORT_COMPLETE_QF_OK; - transport_handle_queue_full(cmd, cmd->se_dev); + + transport_handle_queue_full(cmd, cmd->se_dev, ret, false); } void target_free_sgl(struct scatterlist *sgl, int nents) @@ -2449,18 +2480,14 @@ transport_generic_new_cmd(struct se_cmd *cmd) spin_unlock_irqrestore(&cmd->t_state_lock, flags); ret = cmd->se_tfo->write_pending(cmd); - if (ret == -EAGAIN || ret == -ENOMEM) + if (ret) goto queue_full; - /* fabric drivers should only return -EAGAIN or -ENOMEM as error */ - WARN_ON(ret); - - return (!ret) ? 0 : TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE; + return 0; queue_full: pr_debug("Handling write_pending QUEUE__FULL: se_cmd: %p\n", cmd); - cmd->t_state = TRANSPORT_COMPLETE_QF_WP; - transport_handle_queue_full(cmd, cmd->se_dev); + transport_handle_queue_full(cmd, cmd->se_dev, ret, true); return 0; } EXPORT_SYMBOL(transport_generic_new_cmd); @@ -2470,10 +2497,10 @@ static void transport_write_pending_qf(struct se_cmd *cmd) int ret; ret = cmd->se_tfo->write_pending(cmd); - if (ret == -EAGAIN || ret == -ENOMEM) { + if (ret) { pr_debug("Handling write_pending QUEUE__FULL: se_cmd: %p\n", cmd); - transport_handle_queue_full(cmd, cmd->se_dev); + transport_handle_queue_full(cmd, cmd->se_dev, ret, true); } } @@ -3011,6 +3038,8 @@ static int __transport_check_aborted_status(struct se_cmd *cmd, int send_status) __releases(&cmd->t_state_lock) __acquires(&cmd->t_state_lock) { + int ret; + assert_spin_locked(&cmd->t_state_lock); WARN_ON_ONCE(!irqs_disabled()); @@ -3034,7 +3063,9 @@ static int __transport_check_aborted_status(struct se_cmd *cmd, int send_status) trace_target_cmd_complete(cmd); spin_unlock_irq(&cmd->t_state_lock); - cmd->se_tfo->queue_status(cmd); + ret = cmd->se_tfo->queue_status(cmd); + if (ret) + transport_handle_queue_full(cmd, cmd->se_dev, ret, false); spin_lock_irq(&cmd->t_state_lock); return 1; @@ -3055,6 +3086,7 @@ EXPORT_SYMBOL(transport_check_aborted_status); void transport_send_task_abort(struct se_cmd *cmd) { unsigned long flags; + int ret; spin_lock_irqsave(&cmd->t_state_lock, flags); if (cmd->se_cmd_flags & (SCF_SENT_CHECK_CONDITION)) { @@ -3090,7 +3122,9 @@ send_abort: cmd->t_task_cdb[0], cmd->tag); trace_target_cmd_complete(cmd); - cmd->se_tfo->queue_status(cmd); + ret = cmd->se_tfo->queue_status(cmd); + if (ret) + transport_handle_queue_full(cmd, cmd->se_dev, ret, false); } static void target_tmr_work(struct work_struct *work) diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h index 2e282461cfa5..730ed3055336 100644 --- a/include/target/target_core_base.h +++ b/include/target/target_core_base.h @@ -117,6 +117,7 @@ enum transport_state_table { TRANSPORT_ISTATE_PROCESSING = 11, TRANSPORT_COMPLETE_QF_WP = 18, TRANSPORT_COMPLETE_QF_OK = 19, + TRANSPORT_COMPLETE_QF_ERR = 20, }; /* Used for struct se_cmd->se_cmd_flags */ -- cgit v1.2.3 From d19c4643a52f0a56a7ccc86b145f207a57f40116 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 29 Mar 2017 00:19:24 -0500 Subject: target: Fix ALUA transition state race between multiple initiators Multiple threads could be writing to alua_access_state at the same time, or there could be multiple STPGs in flight (different initiators sending them or one initiator sending them to different ports), or a combo of both and the core_alua_do_transition_tg_pt calls will race with each other. Because from the last patches we no longer delay running core_alua_do_transition_tg_pt_work, there does not seem to be any point in running that in a workqueue. And, we always wait for it to complete one way or another, so we can sleep in this code path. So, this patch made over target-pending just adds a mutex and does the work core_alua_do_transition_tg_pt_work was doing in core_alua_do_transition_tg_pt. There is also no need to use an atomic for the tg_pt_gp_alua_access_state. In core_alua_do_transition_tg_pt we will test and set it under the transition mutex. And, it is a int/32 bits so in the other places where it is read, we will never see it partially updated. Signed-off-by: Mike Christie Signed-off-by: Nicholas Bellinger --- drivers/target/target_core_alua.c | 136 ++++++++++++---------------------- drivers/target/target_core_configfs.c | 2 +- include/target/target_core_base.h | 8 +- 3 files changed, 51 insertions(+), 95 deletions(-) (limited to 'include') diff --git a/drivers/target/target_core_alua.c b/drivers/target/target_core_alua.c index fd7c16a7ca6e..fc4a9c303d55 100644 --- a/drivers/target/target_core_alua.c +++ b/drivers/target/target_core_alua.c @@ -197,8 +197,7 @@ target_emulate_report_target_port_groups(struct se_cmd *cmd) /* * Set the ASYMMETRIC ACCESS State */ - buf[off++] |= (atomic_read( - &tg_pt_gp->tg_pt_gp_alua_access_state) & 0xff); + buf[off++] |= tg_pt_gp->tg_pt_gp_alua_access_state & 0xff; /* * Set supported ASYMMETRIC ACCESS State bits */ @@ -710,7 +709,7 @@ target_alua_state_check(struct se_cmd *cmd) spin_lock(&lun->lun_tg_pt_gp_lock); tg_pt_gp = lun->lun_tg_pt_gp; - out_alua_state = atomic_read(&tg_pt_gp->tg_pt_gp_alua_access_state); + out_alua_state = tg_pt_gp->tg_pt_gp_alua_access_state; nonop_delay_msecs = tg_pt_gp->tg_pt_gp_nonop_delay_msecs; // XXX: keeps using tg_pt_gp witout reference after unlock @@ -911,7 +910,7 @@ static int core_alua_write_tpg_metadata( } /* - * Called with tg_pt_gp->tg_pt_gp_md_mutex held + * Called with tg_pt_gp->tg_pt_gp_transition_mutex held */ static int core_alua_update_tpg_primary_metadata( struct t10_alua_tg_pt_gp *tg_pt_gp) @@ -934,7 +933,7 @@ static int core_alua_update_tpg_primary_metadata( "alua_access_state=0x%02x\n" "alua_access_status=0x%02x\n", tg_pt_gp->tg_pt_gp_id, - tg_pt_gp->tg_pt_gp_alua_pending_state, + tg_pt_gp->tg_pt_gp_alua_access_state, tg_pt_gp->tg_pt_gp_alua_access_status); snprintf(path, ALUA_METADATA_PATH_LEN, @@ -1013,93 +1012,41 @@ static void core_alua_queue_state_change_ua(struct t10_alua_tg_pt_gp *tg_pt_gp) spin_unlock(&tg_pt_gp->tg_pt_gp_lock); } -static void core_alua_do_transition_tg_pt_work(struct work_struct *work) -{ - struct t10_alua_tg_pt_gp *tg_pt_gp = container_of(work, - struct t10_alua_tg_pt_gp, tg_pt_gp_transition_work); - struct se_device *dev = tg_pt_gp->tg_pt_gp_dev; - bool explicit = (tg_pt_gp->tg_pt_gp_alua_access_status == - ALUA_STATUS_ALTERED_BY_EXPLICIT_STPG); - - /* - * Update the ALUA metadata buf that has been allocated in - * core_alua_do_port_transition(), this metadata will be written - * to struct file. - * - * Note that there is the case where we do not want to update the - * metadata when the saved metadata is being parsed in userspace - * when setting the existing port access state and access status. - * - * Also note that the failure to write out the ALUA metadata to - * struct file does NOT affect the actual ALUA transition. - */ - if (tg_pt_gp->tg_pt_gp_write_metadata) { - mutex_lock(&tg_pt_gp->tg_pt_gp_md_mutex); - core_alua_update_tpg_primary_metadata(tg_pt_gp); - mutex_unlock(&tg_pt_gp->tg_pt_gp_md_mutex); - } - /* - * Set the current primary ALUA access state to the requested new state - */ - atomic_set(&tg_pt_gp->tg_pt_gp_alua_access_state, - tg_pt_gp->tg_pt_gp_alua_pending_state); - - pr_debug("Successful %s ALUA transition TG PT Group: %s ID: %hu" - " from primary access state %s to %s\n", (explicit) ? "explicit" : - "implicit", config_item_name(&tg_pt_gp->tg_pt_gp_group.cg_item), - tg_pt_gp->tg_pt_gp_id, - core_alua_dump_state(tg_pt_gp->tg_pt_gp_alua_previous_state), - core_alua_dump_state(tg_pt_gp->tg_pt_gp_alua_pending_state)); - - core_alua_queue_state_change_ua(tg_pt_gp); - - spin_lock(&dev->t10_alua.tg_pt_gps_lock); - atomic_dec(&tg_pt_gp->tg_pt_gp_ref_cnt); - spin_unlock(&dev->t10_alua.tg_pt_gps_lock); - - if (tg_pt_gp->tg_pt_gp_transition_complete) - complete(tg_pt_gp->tg_pt_gp_transition_complete); -} - static int core_alua_do_transition_tg_pt( struct t10_alua_tg_pt_gp *tg_pt_gp, int new_state, int explicit) { - struct se_device *dev = tg_pt_gp->tg_pt_gp_dev; - DECLARE_COMPLETION_ONSTACK(wait); + int prev_state; + mutex_lock(&tg_pt_gp->tg_pt_gp_transition_mutex); /* Nothing to be done here */ - if (atomic_read(&tg_pt_gp->tg_pt_gp_alua_access_state) == new_state) + if (tg_pt_gp->tg_pt_gp_alua_access_state == new_state) { + mutex_unlock(&tg_pt_gp->tg_pt_gp_transition_mutex); return 0; + } - if (explicit && new_state == ALUA_ACCESS_STATE_TRANSITION) + if (explicit && new_state == ALUA_ACCESS_STATE_TRANSITION) { + mutex_unlock(&tg_pt_gp->tg_pt_gp_transition_mutex); return -EAGAIN; - - /* - * Flush any pending transitions - */ - if (!explicit) - flush_work(&tg_pt_gp->tg_pt_gp_transition_work); + } /* * Save the old primary ALUA access state, and set the current state * to ALUA_ACCESS_STATE_TRANSITION. */ - atomic_set(&tg_pt_gp->tg_pt_gp_alua_access_state, - ALUA_ACCESS_STATE_TRANSITION); + prev_state = tg_pt_gp->tg_pt_gp_alua_access_state; + tg_pt_gp->tg_pt_gp_alua_access_state = ALUA_ACCESS_STATE_TRANSITION; tg_pt_gp->tg_pt_gp_alua_access_status = (explicit) ? ALUA_STATUS_ALTERED_BY_EXPLICIT_STPG : ALUA_STATUS_ALTERED_BY_IMPLICIT_ALUA; core_alua_queue_state_change_ua(tg_pt_gp); - if (new_state == ALUA_ACCESS_STATE_TRANSITION) + if (new_state == ALUA_ACCESS_STATE_TRANSITION) { + mutex_unlock(&tg_pt_gp->tg_pt_gp_transition_mutex); return 0; - - tg_pt_gp->tg_pt_gp_alua_previous_state = - atomic_read(&tg_pt_gp->tg_pt_gp_alua_access_state); - tg_pt_gp->tg_pt_gp_alua_pending_state = new_state; + } /* * Check for the optional ALUA primary state transition delay @@ -1108,19 +1055,36 @@ static int core_alua_do_transition_tg_pt( msleep_interruptible(tg_pt_gp->tg_pt_gp_trans_delay_msecs); /* - * Take a reference for workqueue item + * Set the current primary ALUA access state to the requested new state */ - spin_lock(&dev->t10_alua.tg_pt_gps_lock); - atomic_inc(&tg_pt_gp->tg_pt_gp_ref_cnt); - spin_unlock(&dev->t10_alua.tg_pt_gps_lock); + tg_pt_gp->tg_pt_gp_alua_access_state = new_state; - schedule_work(&tg_pt_gp->tg_pt_gp_transition_work); - if (explicit) { - tg_pt_gp->tg_pt_gp_transition_complete = &wait; - wait_for_completion(&wait); - tg_pt_gp->tg_pt_gp_transition_complete = NULL; + /* + * Update the ALUA metadata buf that has been allocated in + * core_alua_do_port_transition(), this metadata will be written + * to struct file. + * + * Note that there is the case where we do not want to update the + * metadata when the saved metadata is being parsed in userspace + * when setting the existing port access state and access status. + * + * Also note that the failure to write out the ALUA metadata to + * struct file does NOT affect the actual ALUA transition. + */ + if (tg_pt_gp->tg_pt_gp_write_metadata) { + core_alua_update_tpg_primary_metadata(tg_pt_gp); } + pr_debug("Successful %s ALUA transition TG PT Group: %s ID: %hu" + " from primary access state %s to %s\n", (explicit) ? "explicit" : + "implicit", config_item_name(&tg_pt_gp->tg_pt_gp_group.cg_item), + tg_pt_gp->tg_pt_gp_id, + core_alua_dump_state(prev_state), + core_alua_dump_state(new_state)); + + core_alua_queue_state_change_ua(tg_pt_gp); + + mutex_unlock(&tg_pt_gp->tg_pt_gp_transition_mutex); return 0; } @@ -1685,14 +1649,12 @@ struct t10_alua_tg_pt_gp *core_alua_allocate_tg_pt_gp(struct se_device *dev, } INIT_LIST_HEAD(&tg_pt_gp->tg_pt_gp_list); INIT_LIST_HEAD(&tg_pt_gp->tg_pt_gp_lun_list); - mutex_init(&tg_pt_gp->tg_pt_gp_md_mutex); + mutex_init(&tg_pt_gp->tg_pt_gp_transition_mutex); spin_lock_init(&tg_pt_gp->tg_pt_gp_lock); atomic_set(&tg_pt_gp->tg_pt_gp_ref_cnt, 0); - INIT_WORK(&tg_pt_gp->tg_pt_gp_transition_work, - core_alua_do_transition_tg_pt_work); tg_pt_gp->tg_pt_gp_dev = dev; - atomic_set(&tg_pt_gp->tg_pt_gp_alua_access_state, - ALUA_ACCESS_STATE_ACTIVE_OPTIMIZED); + tg_pt_gp->tg_pt_gp_alua_access_state = + ALUA_ACCESS_STATE_ACTIVE_OPTIMIZED; /* * Enable both explicit and implicit ALUA support by default */ @@ -1797,8 +1759,6 @@ void core_alua_free_tg_pt_gp( dev->t10_alua.alua_tg_pt_gps_counter--; spin_unlock(&dev->t10_alua.tg_pt_gps_lock); - flush_work(&tg_pt_gp->tg_pt_gp_transition_work); - /* * Allow a struct t10_alua_tg_pt_gp_member * referenced by * core_alua_get_tg_pt_gp_by_name() in @@ -1938,8 +1898,8 @@ ssize_t core_alua_show_tg_pt_gp_info(struct se_lun *lun, char *page) "Primary Access Status: %s\nTG Port Secondary Access" " State: %s\nTG Port Secondary Access Status: %s\n", config_item_name(tg_pt_ci), tg_pt_gp->tg_pt_gp_id, - core_alua_dump_state(atomic_read( - &tg_pt_gp->tg_pt_gp_alua_access_state)), + core_alua_dump_state( + tg_pt_gp->tg_pt_gp_alua_access_state), core_alua_dump_status( tg_pt_gp->tg_pt_gp_alua_access_status), atomic_read(&lun->lun_tg_pt_secondary_offline) ? diff --git a/drivers/target/target_core_configfs.c b/drivers/target/target_core_configfs.c index 38b5025e4c7a..70657fd56440 100644 --- a/drivers/target/target_core_configfs.c +++ b/drivers/target/target_core_configfs.c @@ -2392,7 +2392,7 @@ static ssize_t target_tg_pt_gp_alua_access_state_show(struct config_item *item, char *page) { return sprintf(page, "%d\n", - atomic_read(&to_tg_pt_gp(item)->tg_pt_gp_alua_access_state)); + to_tg_pt_gp(item)->tg_pt_gp_alua_access_state); } static ssize_t target_tg_pt_gp_alua_access_state_store(struct config_item *item, diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h index 730ed3055336..ccfad0e9c2cd 100644 --- a/include/target/target_core_base.h +++ b/include/target/target_core_base.h @@ -280,8 +280,6 @@ struct t10_alua_tg_pt_gp { u16 tg_pt_gp_id; int tg_pt_gp_valid_id; int tg_pt_gp_alua_supported_states; - int tg_pt_gp_alua_pending_state; - int tg_pt_gp_alua_previous_state; int tg_pt_gp_alua_access_status; int tg_pt_gp_alua_access_type; int tg_pt_gp_nonop_delay_msecs; @@ -290,18 +288,16 @@ struct t10_alua_tg_pt_gp { int tg_pt_gp_pref; int tg_pt_gp_write_metadata; u32 tg_pt_gp_members; - atomic_t tg_pt_gp_alua_access_state; + int tg_pt_gp_alua_access_state; atomic_t tg_pt_gp_ref_cnt; spinlock_t tg_pt_gp_lock; - struct mutex tg_pt_gp_md_mutex; + struct mutex tg_pt_gp_transition_mutex; struct se_device *tg_pt_gp_dev; struct config_group tg_pt_gp_group; struct list_head tg_pt_gp_list; struct list_head tg_pt_gp_lun_list; struct se_lun *tg_pt_gp_alua_lun; struct se_node_acl *tg_pt_gp_alua_nacl; - struct work_struct tg_pt_gp_transition_work; - struct completion *tg_pt_gp_transition_complete; }; struct t10_vpd { -- cgit v1.2.3 From 27c0e3748e41ca79171ffa3e97415a20af6facd0 Mon Sep 17 00:00:00 2001 From: Al Viro Date: Fri, 17 Feb 2017 18:42:24 -0500 Subject: [iov_iter] new privimitive: iov_iter_revert() opposite to iov_iter_advance(); the caller is responsible for never using it to move back past the initial position. Cc: stable@vger.kernel.org Signed-off-by: Al Viro --- include/linux/uio.h | 6 ++++- lib/iov_iter.c | 63 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 68 insertions(+), 1 deletion(-) (limited to 'include') diff --git a/include/linux/uio.h b/include/linux/uio.h index 804e34c6f981..f2d36a3d3005 100644 --- a/include/linux/uio.h +++ b/include/linux/uio.h @@ -39,7 +39,10 @@ struct iov_iter { }; union { unsigned long nr_segs; - int idx; + struct { + int idx; + int start_idx; + }; }; }; @@ -81,6 +84,7 @@ unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to); size_t iov_iter_copy_from_user_atomic(struct page *page, struct iov_iter *i, unsigned long offset, size_t bytes); void iov_iter_advance(struct iov_iter *i, size_t bytes); +void iov_iter_revert(struct iov_iter *i, size_t bytes); int iov_iter_fault_in_readable(struct iov_iter *i, size_t bytes); size_t iov_iter_single_seg_count(const struct iov_iter *i); size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes, diff --git a/lib/iov_iter.c b/lib/iov_iter.c index e68604ae3ced..60abc44385b7 100644 --- a/lib/iov_iter.c +++ b/lib/iov_iter.c @@ -786,6 +786,68 @@ void iov_iter_advance(struct iov_iter *i, size_t size) } EXPORT_SYMBOL(iov_iter_advance); +void iov_iter_revert(struct iov_iter *i, size_t unroll) +{ + if (!unroll) + return; + i->count += unroll; + if (unlikely(i->type & ITER_PIPE)) { + struct pipe_inode_info *pipe = i->pipe; + int idx = i->idx; + size_t off = i->iov_offset; + while (1) { + size_t n = off - pipe->bufs[idx].offset; + if (unroll < n) { + off -= (n - unroll); + break; + } + unroll -= n; + if (!unroll && idx == i->start_idx) { + off = 0; + break; + } + if (!idx--) + idx = pipe->buffers - 1; + off = pipe->bufs[idx].offset + pipe->bufs[idx].len; + } + i->iov_offset = off; + i->idx = idx; + pipe_truncate(i); + return; + } + if (unroll <= i->iov_offset) { + i->iov_offset -= unroll; + return; + } + unroll -= i->iov_offset; + if (i->type & ITER_BVEC) { + const struct bio_vec *bvec = i->bvec; + while (1) { + size_t n = (--bvec)->bv_len; + i->nr_segs++; + if (unroll <= n) { + i->bvec = bvec; + i->iov_offset = n - unroll; + return; + } + unroll -= n; + } + } else { /* same logics for iovec and kvec */ + const struct iovec *iov = i->iov; + while (1) { + size_t n = (--iov)->iov_len; + i->nr_segs++; + if (unroll <= n) { + i->iov = iov; + i->iov_offset = n - unroll; + return; + } + unroll -= n; + } + } +} +EXPORT_SYMBOL(iov_iter_revert); + /* * Return the count of just the current iov_iter segment. */ @@ -839,6 +901,7 @@ void iov_iter_pipe(struct iov_iter *i, int direction, i->idx = (pipe->curbuf + pipe->nrbufs) & (pipe->buffers - 1); i->iov_offset = 0; i->count = count; + i->start_idx = i->idx; } EXPORT_SYMBOL(iov_iter_pipe); -- cgit v1.2.3 From 62e24c5775ecb387a3eb33701378ccfa6dbc98ee Mon Sep 17 00:00:00 2001 From: Philipp Zabel Date: Fri, 5 Feb 2016 13:41:39 +0100 Subject: reset: add exported __reset_control_get, return NULL if optional Rename the internal __reset_control_get/put functions to __reset_control_get/put_internal and add an exported __reset_control_get equivalent to __of_reset_control_get that takes a struct device parameter. This avoids the confusing call to __of_reset_control_get in the non-DT case and fixes the devm_reset_control_get_optional function to return NULL if RESET_CONTROLLER is enabled but dev->of_node == NULL. Fixes: bb475230b8e5 ("reset: make optional functions really optional") Reported-by: Andy Shevchenko Tested-by: Andy Shevchenko Cc: Ramiro Oliveira Signed-off-by: Philipp Zabel --- drivers/reset/core.c | 22 ++++++++++++++++------ include/linux/reset.h | 22 ++++++++++++++-------- 2 files changed, 30 insertions(+), 14 deletions(-) (limited to 'include') diff --git a/drivers/reset/core.c b/drivers/reset/core.c index f1e5e65388bb..cd739d2fa160 100644 --- a/drivers/reset/core.c +++ b/drivers/reset/core.c @@ -275,7 +275,7 @@ int reset_control_status(struct reset_control *rstc) } EXPORT_SYMBOL_GPL(reset_control_status); -static struct reset_control *__reset_control_get( +static struct reset_control *__reset_control_get_internal( struct reset_controller_dev *rcdev, unsigned int index, bool shared) { @@ -308,7 +308,7 @@ static struct reset_control *__reset_control_get( return rstc; } -static void __reset_control_put(struct reset_control *rstc) +static void __reset_control_put_internal(struct reset_control *rstc) { lockdep_assert_held(&reset_list_mutex); @@ -377,7 +377,7 @@ struct reset_control *__of_reset_control_get(struct device_node *node, } /* reset_list_mutex also protects the rcdev's reset_control list */ - rstc = __reset_control_get(rcdev, rstc_id, shared); + rstc = __reset_control_get_internal(rcdev, rstc_id, shared); mutex_unlock(&reset_list_mutex); @@ -385,6 +385,17 @@ struct reset_control *__of_reset_control_get(struct device_node *node, } EXPORT_SYMBOL_GPL(__of_reset_control_get); +struct reset_control *__reset_control_get(struct device *dev, const char *id, + int index, bool shared, bool optional) +{ + if (dev->of_node) + return __of_reset_control_get(dev->of_node, id, index, shared, + optional); + + return optional ? NULL : ERR_PTR(-EINVAL); +} +EXPORT_SYMBOL_GPL(__reset_control_get); + /** * reset_control_put - free the reset controller * @rstc: reset controller @@ -396,7 +407,7 @@ void reset_control_put(struct reset_control *rstc) return; mutex_lock(&reset_list_mutex); - __reset_control_put(rstc); + __reset_control_put_internal(rstc); mutex_unlock(&reset_list_mutex); } EXPORT_SYMBOL_GPL(reset_control_put); @@ -417,8 +428,7 @@ struct reset_control *__devm_reset_control_get(struct device *dev, if (!ptr) return ERR_PTR(-ENOMEM); - rstc = __of_reset_control_get(dev ? dev->of_node : NULL, - id, index, shared, optional); + rstc = __reset_control_get(dev, id, index, shared, optional); if (!IS_ERR(rstc)) { *ptr = rstc; devres_add(dev, ptr); diff --git a/include/linux/reset.h b/include/linux/reset.h index 96fb139bdd08..13d8681210d5 100644 --- a/include/linux/reset.h +++ b/include/linux/reset.h @@ -15,6 +15,9 @@ int reset_control_status(struct reset_control *rstc); struct reset_control *__of_reset_control_get(struct device_node *node, const char *id, int index, bool shared, bool optional); +struct reset_control *__reset_control_get(struct device *dev, const char *id, + int index, bool shared, + bool optional); void reset_control_put(struct reset_control *rstc); struct reset_control *__devm_reset_control_get(struct device *dev, const char *id, int index, bool shared, @@ -72,6 +75,13 @@ static inline struct reset_control *__of_reset_control_get( return optional ? NULL : ERR_PTR(-ENOTSUPP); } +static inline struct reset_control *__reset_control_get( + struct device *dev, const char *id, + int index, bool shared, bool optional) +{ + return optional ? NULL : ERR_PTR(-ENOTSUPP); +} + static inline struct reset_control *__devm_reset_control_get( struct device *dev, const char *id, int index, bool shared, bool optional) @@ -102,8 +112,7 @@ __must_check reset_control_get_exclusive(struct device *dev, const char *id) #ifndef CONFIG_RESET_CONTROLLER WARN_ON(1); #endif - return __of_reset_control_get(dev ? dev->of_node : NULL, id, 0, false, - false); + return __reset_control_get(dev, id, 0, false, false); } /** @@ -131,22 +140,19 @@ __must_check reset_control_get_exclusive(struct device *dev, const char *id) static inline struct reset_control *reset_control_get_shared( struct device *dev, const char *id) { - return __of_reset_control_get(dev ? dev->of_node : NULL, id, 0, true, - false); + return __reset_control_get(dev, id, 0, true, false); } static inline struct reset_control *reset_control_get_optional_exclusive( struct device *dev, const char *id) { - return __of_reset_control_get(dev ? dev->of_node : NULL, id, 0, false, - true); + return __reset_control_get(dev, id, 0, false, true); } static inline struct reset_control *reset_control_get_optional_shared( struct device *dev, const char *id) { - return __of_reset_control_get(dev ? dev->of_node : NULL, id, 0, true, - true); + return __reset_control_get(dev, id, 0, true, true); } /** -- cgit v1.2.3 From 404123c2db798027e852480ed9c4accef9f1d9e6 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Wed, 29 Mar 2017 19:06:20 +0300 Subject: virtio: allow drivers to validate features Some drivers can't support all features in all configurations. At the moment we blindly set FEATURES_OK and later FAILED. Support this better by adding a callback drivers can use to do some early checks. Signed-off-by: Michael S. Tsirkin --- drivers/virtio/virtio.c | 6 ++++++ include/linux/virtio.h | 1 + 2 files changed, 7 insertions(+) (limited to 'include') diff --git a/drivers/virtio/virtio.c b/drivers/virtio/virtio.c index 400d70b69379..48230a5e12f2 100644 --- a/drivers/virtio/virtio.c +++ b/drivers/virtio/virtio.c @@ -232,6 +232,12 @@ static int virtio_dev_probe(struct device *_d) if (device_features & (1ULL << i)) __virtio_set_bit(dev, i); + if (drv->validate) { + err = drv->validate(dev); + if (err) + goto err; + } + err = virtio_finalize_features(dev); if (err) goto err; diff --git a/include/linux/virtio.h b/include/linux/virtio.h index 04b0d3f95043..7edfbdb55a99 100644 --- a/include/linux/virtio.h +++ b/include/linux/virtio.h @@ -167,6 +167,7 @@ struct virtio_driver { unsigned int feature_table_size; const unsigned int *feature_table_legacy; unsigned int feature_table_size_legacy; + int (*validate)(struct virtio_device *dev); int (*probe)(struct virtio_device *dev); void (*scan)(struct virtio_device *dev); void (*remove)(struct virtio_device *dev); -- cgit v1.2.3 From 2008c1544c73d5190f81ef1790fa5bd2fade5bd0 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Tue, 4 Apr 2017 21:09:20 +0300 Subject: Revert "virtio_pci: don't duplicate the msix_enable flag in struct pci_dev" This reverts commit 53a020c661741f3b87ad3ac6fa545088aaebac9b. The cleanup seems to be one of the changes that broke hybernation for some users. We are still not sure why but revert helps. Tested-by: Mike Galbraith Signed-off-by: Michael S. Tsirkin --- drivers/virtio/virtio_pci_common.c | 5 +++-- drivers/virtio/virtio_pci_common.h | 2 ++ drivers/virtio/virtio_pci_legacy.c | 2 +- drivers/virtio/virtio_pci_modern.c | 2 +- include/uapi/linux/virtio_pci.h | 2 +- 5 files changed, 8 insertions(+), 5 deletions(-) (limited to 'include') diff --git a/drivers/virtio/virtio_pci_common.c b/drivers/virtio/virtio_pci_common.c index 4608fd9aaa6c..3921b0a2439e 100644 --- a/drivers/virtio/virtio_pci_common.c +++ b/drivers/virtio/virtio_pci_common.c @@ -125,7 +125,7 @@ void vp_del_vqs(struct virtio_device *vdev) vp_remove_vqs(vdev); - if (vp_dev->pci_dev->msix_enabled) { + if (vp_dev->msix_enabled) { for (i = 0; i < vp_dev->msix_vectors; i++) free_cpumask_var(vp_dev->msix_affinity_masks[i]); @@ -249,6 +249,7 @@ static int vp_find_vqs_msix(struct virtio_device *vdev, unsigned nvqs, allocated_vectors++; } + vp_dev->msix_enabled = 1; return 0; out_remove_vqs: @@ -343,7 +344,7 @@ int vp_set_vq_affinity(struct virtqueue *vq, int cpu) if (!vq->callback) return -EINVAL; - if (vp_dev->pci_dev->msix_enabled) { + if (vp_dev->msix_enabled) { int vec = vp_dev->msix_vector_map[vq->index]; struct cpumask *mask = vp_dev->msix_affinity_masks[vec]; unsigned int irq = pci_irq_vector(vp_dev->pci_dev, vec); diff --git a/drivers/virtio/virtio_pci_common.h b/drivers/virtio/virtio_pci_common.h index ac8c9d788964..c8074997fd28 100644 --- a/drivers/virtio/virtio_pci_common.h +++ b/drivers/virtio/virtio_pci_common.h @@ -64,6 +64,8 @@ struct virtio_pci_device { /* the IO mapping for the PCI config space */ void __iomem *ioaddr; + /* MSI-X support */ + int msix_enabled; cpumask_var_t *msix_affinity_masks; /* Name strings for interrupts. This size should be enough, * and I'm too lazy to allocate each name separately. */ diff --git a/drivers/virtio/virtio_pci_legacy.c b/drivers/virtio/virtio_pci_legacy.c index f7362c5fe18a..5dd01f09608b 100644 --- a/drivers/virtio/virtio_pci_legacy.c +++ b/drivers/virtio/virtio_pci_legacy.c @@ -165,7 +165,7 @@ static void del_vq(struct virtqueue *vq) iowrite16(vq->index, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_SEL); - if (vp_dev->pci_dev->msix_enabled) { + if (vp_dev->msix_enabled) { iowrite16(VIRTIO_MSI_NO_VECTOR, vp_dev->ioaddr + VIRTIO_MSI_QUEUE_VECTOR); /* Flush the write out to device */ diff --git a/drivers/virtio/virtio_pci_modern.c b/drivers/virtio/virtio_pci_modern.c index 7bc3004b840e..7ce36daccc31 100644 --- a/drivers/virtio/virtio_pci_modern.c +++ b/drivers/virtio/virtio_pci_modern.c @@ -411,7 +411,7 @@ static void del_vq(struct virtqueue *vq) vp_iowrite16(vq->index, &vp_dev->common->queue_select); - if (vp_dev->pci_dev->msix_enabled) { + if (vp_dev->msix_enabled) { vp_iowrite16(VIRTIO_MSI_NO_VECTOR, &vp_dev->common->queue_msix_vector); /* Flush the write out to device */ diff --git a/include/uapi/linux/virtio_pci.h b/include/uapi/linux/virtio_pci.h index 15b4385a2be1..90007a1abcab 100644 --- a/include/uapi/linux/virtio_pci.h +++ b/include/uapi/linux/virtio_pci.h @@ -79,7 +79,7 @@ * configuration space */ #define VIRTIO_PCI_CONFIG_OFF(msix_enabled) ((msix_enabled) ? 24 : 20) /* Deprecated: please use VIRTIO_PCI_CONFIG_OFF instead */ -#define VIRTIO_PCI_CONFIG(dev) VIRTIO_PCI_CONFIG_OFF((dev)->pci_dev->msix_enabled) +#define VIRTIO_PCI_CONFIG(dev) VIRTIO_PCI_CONFIG_OFF((dev)->msix_enabled) /* Virtio ABI version, this must match exactly */ #define VIRTIO_PCI_ABI_VERSION 0 -- cgit v1.2.3 From c0c379e2931b05facef538e53bf3b21f283d9a0b Mon Sep 17 00:00:00 2001 From: "Kirill A. Shutemov" Date: Thu, 13 Apr 2017 14:56:23 -0700 Subject: mm: drop unused pmdp_huge_get_and_clear_notify() Dave noticed that after fixing MADV_DONTNEED vs numa balancing race the last pmdp_huge_get_and_clear_notify() user is gone. Let's drop the helper. Link: http://lkml.kernel.org/r/20170306112047.24809-1-kirill.shutemov@linux.intel.com Signed-off-by: Kirill A. Shutemov Cc: Dave Hansen Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/mmu_notifier.h | 13 ------------- 1 file changed, 13 deletions(-) (limited to 'include') diff --git a/include/linux/mmu_notifier.h b/include/linux/mmu_notifier.h index 51891fb0d3ce..c91b3bcd158f 100644 --- a/include/linux/mmu_notifier.h +++ b/include/linux/mmu_notifier.h @@ -394,18 +394,6 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm) ___pud; \ }) -#define pmdp_huge_get_and_clear_notify(__mm, __haddr, __pmd) \ -({ \ - unsigned long ___haddr = __haddr & HPAGE_PMD_MASK; \ - pmd_t ___pmd; \ - \ - ___pmd = pmdp_huge_get_and_clear(__mm, __haddr, __pmd); \ - mmu_notifier_invalidate_range(__mm, ___haddr, \ - ___haddr + HPAGE_PMD_SIZE); \ - \ - ___pmd; \ -}) - /* * set_pte_at_notify() sets the pte _after_ running the notifier. * This is safe to start by updating the secondary MMUs, because the primary MMU @@ -489,7 +477,6 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm) #define ptep_clear_flush_notify ptep_clear_flush #define pmdp_huge_clear_flush_notify pmdp_huge_clear_flush #define pudp_huge_clear_flush_notify pudp_huge_clear_flush -#define pmdp_huge_get_and_clear_notify pmdp_huge_get_and_clear #define set_pte_at_notify set_pte_at #endif /* CONFIG_MMU_NOTIFIER */ -- cgit v1.2.3 From 5a8d75a1b8c99bdc926ba69b7b7dbe4fae81a5af Mon Sep 17 00:00:00 2001 From: Ming Lei Date: Fri, 14 Apr 2017 13:58:29 -0600 Subject: block: fix bio_will_gap() for first bvec with offset Commit 729204ef49ec("block: relax check on sg gap") allows us to merge bios, if both are physically contiguous. This change can merge a huge number of small bios, through mkfs for example, mkfs.ntfs running time can be decreased to ~1/10. But if one rq starts with a non-aligned buffer (the 1st bvec's bv_offset is non-zero) and if we allow the merge, it is quite difficult to respect sg gap limit, especially the max segment size, or we risk having an unaligned virtual boundary. This patch tries to avoid the issue by disallowing a merge, if the req starts with an unaligned buffer. Also add comments to explain why the merged segment can't end in unaligned virt boundary. Fixes: 729204ef49ec ("block: relax check on sg gap") Tested-by: Johannes Thumshirn Reviewed-by: Johannes Thumshirn Signed-off-by: Ming Lei Rewrote parts of the commit message and comments. Signed-off-by: Jens Axboe --- include/linux/blkdev.h | 32 ++++++++++++++++++++++++++++---- 1 file changed, 28 insertions(+), 4 deletions(-) (limited to 'include') diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 7548f332121a..01a696b0a4d3 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -1672,12 +1672,36 @@ static inline bool bios_segs_mergeable(struct request_queue *q, return true; } -static inline bool bio_will_gap(struct request_queue *q, struct bio *prev, - struct bio *next) +static inline bool bio_will_gap(struct request_queue *q, + struct request *prev_rq, + struct bio *prev, + struct bio *next) { if (bio_has_data(prev) && queue_virt_boundary(q)) { struct bio_vec pb, nb; + /* + * don't merge if the 1st bio starts with non-zero + * offset, otherwise it is quite difficult to respect + * sg gap limit. We work hard to merge a huge number of small + * single bios in case of mkfs. + */ + if (prev_rq) + bio_get_first_bvec(prev_rq->bio, &pb); + else + bio_get_first_bvec(prev, &pb); + if (pb.bv_offset) + return true; + + /* + * We don't need to worry about the situation that the + * merged segment ends in unaligned virt boundary: + * + * - if 'pb' ends aligned, the merged segment ends aligned + * - if 'pb' ends unaligned, the next bio must include + * one single bvec of 'nb', otherwise the 'nb' can't + * merge with 'pb' + */ bio_get_last_bvec(prev, &pb); bio_get_first_bvec(next, &nb); @@ -1690,12 +1714,12 @@ static inline bool bio_will_gap(struct request_queue *q, struct bio *prev, static inline bool req_gap_back_merge(struct request *req, struct bio *bio) { - return bio_will_gap(req->q, req->biotail, bio); + return bio_will_gap(req->q, req, req->biotail, bio); } static inline bool req_gap_front_merge(struct request *req, struct bio *bio) { - return bio_will_gap(req->q, bio, req->bio); + return bio_will_gap(req->q, NULL, bio, req->bio); } int kblockd_schedule_work(struct work_struct *work); -- cgit v1.2.3