From a7b8829d242b1a58107e9c02b09e93aec446d55c Mon Sep 17 00:00:00 2001 From: Lorenzo Bianconi Date: Wed, 5 Jul 2017 20:30:01 +0200 Subject: iio: accel: st_accel: add SPI-3wire support Add SPI Serial Interface Mode (SIM) register information in st_sensor_settings look up table to support devices (like LSM303AGR accel sensor) that allow just SPI-3wire communication mode. SIM mode has to be configured before any other operation since it is not enabled by default and the driver is not able to read without that configuration Whilst a fairly substantial patch, the actual logic is simple and it is better to have the generic fix than a band aid. Fixes: ddc05fa28606 (iio: st-accel: add support for lsm303agr accel) Signed-off-by: Lorenzo Bianconi Cc: Signed-off-by: Jonathan Cameron --- include/linux/iio/common/st_sensors.h | 7 +++++++ include/linux/platform_data/st_sensors_pdata.h | 2 ++ 2 files changed, 9 insertions(+) (limited to 'include/linux') diff --git a/include/linux/iio/common/st_sensors.h b/include/linux/iio/common/st_sensors.h index 497f2b3a5a62..97f1b465d04f 100644 --- a/include/linux/iio/common/st_sensors.h +++ b/include/linux/iio/common/st_sensors.h @@ -105,6 +105,11 @@ struct st_sensor_fullscale { struct st_sensor_fullscale_avl fs_avl[ST_SENSORS_FULLSCALE_AVL_MAX]; }; +struct st_sensor_sim { + u8 addr; + u8 value; +}; + /** * struct st_sensor_bdu - ST sensor device block data update * @addr: address of the register. @@ -197,6 +202,7 @@ struct st_sensor_transfer_function { * @bdu: Block data update register. * @das: Data Alignment Selection register. * @drdy_irq: Data ready register of the sensor. + * @sim: SPI serial interface mode register of the sensor. * @multi_read_bit: Use or not particular bit for [I2C/SPI] multi-read. * @bootime: samples to discard when sensor passing from power-down to power-up. */ @@ -213,6 +219,7 @@ struct st_sensor_settings { struct st_sensor_bdu bdu; struct st_sensor_das das; struct st_sensor_data_ready_irq drdy_irq; + struct st_sensor_sim sim; bool multi_read_bit; unsigned int bootime; }; diff --git a/include/linux/platform_data/st_sensors_pdata.h b/include/linux/platform_data/st_sensors_pdata.h index 79b0e4cdb814..f8274b0c6888 100644 --- a/include/linux/platform_data/st_sensors_pdata.h +++ b/include/linux/platform_data/st_sensors_pdata.h @@ -17,10 +17,12 @@ * Available only for accelerometer and pressure sensors. * Accelerometer DRDY on LSM330 available only on pin 1 (see datasheet). * @open_drain: set the interrupt line to be open drain if possible. + * @spi_3wire: enable spi-3wire mode. */ struct st_sensors_platform_data { u8 drdy_int_pin; bool open_drain; + bool spi_3wire; }; #endif /* ST_SENSORS_PDATA_H */ -- cgit v1.2.3 From 1937f8a29f4a650bc27e0311b43b53509a34fd22 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Mon, 24 Jul 2017 12:52:59 +0200 Subject: scsi: bnx2fc: Simplify CPU hotplug code The CPU hotplug related code of this driver can be simplified by: 1) Consolidating the callbacks into a single state. The CPU thread can be torn down on the CPU which goes offline. There is no point in delaying that to the CPU dead state 2) Let the core code invoke the online/offline callbacks and remove the extra for_each_online_cpu() loops. Signed-off-by: Thomas Gleixner Signed-off-by: Martin K. Petersen --- drivers/scsi/bnx2fc/bnx2fc_fcoe.c | 69 +++++++++------------------------------ include/linux/cpuhotplug.h | 1 - 2 files changed, 15 insertions(+), 55 deletions(-) (limited to 'include/linux') diff --git a/drivers/scsi/bnx2fc/bnx2fc_fcoe.c b/drivers/scsi/bnx2fc/bnx2fc_fcoe.c index c4ebf8c5d884..6844ba361616 100644 --- a/drivers/scsi/bnx2fc/bnx2fc_fcoe.c +++ b/drivers/scsi/bnx2fc/bnx2fc_fcoe.c @@ -2624,12 +2624,11 @@ static struct fcoe_transport bnx2fc_transport = { }; /** - * bnx2fc_percpu_thread_create - Create a receive thread for an - * online CPU + * bnx2fc_cpu_online - Create a receive thread for an online CPU * * @cpu: cpu index for the online cpu */ -static void bnx2fc_percpu_thread_create(unsigned int cpu) +static int bnx2fc_cpu_online(unsigned int cpu) { struct bnx2fc_percpu_s *p; struct task_struct *thread; @@ -2639,15 +2638,17 @@ static void bnx2fc_percpu_thread_create(unsigned int cpu) thread = kthread_create_on_node(bnx2fc_percpu_io_thread, (void *)p, cpu_to_node(cpu), "bnx2fc_thread/%d", cpu); + if (IS_ERR(thread)) + return PTR_ERR(thread); + /* bind thread to the cpu */ - if (likely(!IS_ERR(thread))) { - kthread_bind(thread, cpu); - p->iothread = thread; - wake_up_process(thread); - } + kthread_bind(thread, cpu); + p->iothread = thread; + wake_up_process(thread); + return 0; } -static void bnx2fc_percpu_thread_destroy(unsigned int cpu) +static int bnx2fc_cpu_offline(unsigned int cpu) { struct bnx2fc_percpu_s *p; struct task_struct *thread; @@ -2661,7 +2662,6 @@ static void bnx2fc_percpu_thread_destroy(unsigned int cpu) thread = p->iothread; p->iothread = NULL; - /* Free all work in the list */ list_for_each_entry_safe(work, tmp, &p->work_list, list) { list_del_init(&work->list); @@ -2673,20 +2673,6 @@ static void bnx2fc_percpu_thread_destroy(unsigned int cpu) if (thread) kthread_stop(thread); -} - - -static int bnx2fc_cpu_online(unsigned int cpu) -{ - printk(PFX "CPU %x online: Create Rx thread\n", cpu); - bnx2fc_percpu_thread_create(cpu); - return 0; -} - -static int bnx2fc_cpu_dead(unsigned int cpu) -{ - printk(PFX "CPU %x offline: Remove Rx thread\n", cpu); - bnx2fc_percpu_thread_destroy(cpu); return 0; } @@ -2761,31 +2747,16 @@ static int __init bnx2fc_mod_init(void) spin_lock_init(&p->fp_work_lock); } - get_online_cpus(); - - for_each_online_cpu(cpu) - bnx2fc_percpu_thread_create(cpu); - - rc = cpuhp_setup_state_nocalls_cpuslocked(CPUHP_AP_ONLINE_DYN, - "scsi/bnx2fc:online", - bnx2fc_cpu_online, NULL); + rc = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "scsi/bnx2fc:online", + bnx2fc_cpu_online, bnx2fc_cpu_offline); if (rc < 0) - goto stop_threads; + goto stop_thread; bnx2fc_online_state = rc; - cpuhp_setup_state_nocalls_cpuslocked(CPUHP_SCSI_BNX2FC_DEAD, - "scsi/bnx2fc:dead", - NULL, bnx2fc_cpu_dead); - put_online_cpus(); - cnic_register_driver(CNIC_ULP_FCOE, &bnx2fc_cnic_cb); - return 0; -stop_threads: - for_each_online_cpu(cpu) - bnx2fc_percpu_thread_destroy(cpu); - put_online_cpus(); +stop_thread: kthread_stop(l2_thread); free_wq: destroy_workqueue(bnx2fc_wq); @@ -2804,7 +2775,6 @@ static void __exit bnx2fc_mod_exit(void) struct fcoe_percpu_s *bg; struct task_struct *l2_thread; struct sk_buff *skb; - unsigned int cpu = 0; /* * NOTE: Since cnic calls register_driver routine rtnl_lock, @@ -2845,16 +2815,7 @@ static void __exit bnx2fc_mod_exit(void) if (l2_thread) kthread_stop(l2_thread); - get_online_cpus(); - /* Destroy per cpu threads */ - for_each_online_cpu(cpu) { - bnx2fc_percpu_thread_destroy(cpu); - } - - cpuhp_remove_state_nocalls_cpuslocked(bnx2fc_online_state); - cpuhp_remove_state_nocalls_cpuslocked(CPUHP_SCSI_BNX2FC_DEAD); - - put_online_cpus(); + cpuhp_remove_state(bnx2fc_online_state); destroy_workqueue(bnx2fc_wq); /* diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h index b56573bf440d..2e7b1731ad12 100644 --- a/include/linux/cpuhotplug.h +++ b/include/linux/cpuhotplug.h @@ -39,7 +39,6 @@ enum cpuhp_state { CPUHP_PCI_XGENE_DEAD, CPUHP_IOMMU_INTEL_DEAD, CPUHP_LUSTRE_CFS_DEAD, - CPUHP_SCSI_BNX2FC_DEAD, CPUHP_SCSI_BNX2I_DEAD, CPUHP_WORKQUEUE_PREP, CPUHP_POWER_NUMA_PREPARE, -- cgit v1.2.3 From f9f22a86912f9d36b50e9b3b383fabfb9f22dd46 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Mon, 24 Jul 2017 12:53:00 +0200 Subject: scsi: bnx2i: Simplify cpu hotplug code The CPU hotplug related code of this driver can be simplified by: 1) Consolidating the callbacks into a single state. The CPU thread can be torn down on the CPU which goes offline. There is no point in delaying that to the CPU dead state 2) Let the core code invoke the online/offline callbacks and remove the extra for_each_online_cpu() loops. Signed-off-by: Thomas Gleixner Acked-by: Chad Dupuis Signed-off-by: Martin K. Petersen --- drivers/scsi/bnx2i/bnx2i_init.c | 65 ++++++++++------------------------------- include/linux/cpuhotplug.h | 1 - 2 files changed, 15 insertions(+), 51 deletions(-) (limited to 'include/linux') diff --git a/drivers/scsi/bnx2i/bnx2i_init.c b/drivers/scsi/bnx2i/bnx2i_init.c index 7487b653e799..4ebcda8d9500 100644 --- a/drivers/scsi/bnx2i/bnx2i_init.c +++ b/drivers/scsi/bnx2i/bnx2i_init.c @@ -404,12 +404,11 @@ int bnx2i_get_stats(void *handle) /** - * bnx2i_percpu_thread_create - Create a receive thread for an - * online CPU + * bnx2i_cpu_online - Create a receive thread for an online CPU * * @cpu: cpu index for the online cpu */ -static void bnx2i_percpu_thread_create(unsigned int cpu) +static int bnx2i_cpu_online(unsigned int cpu) { struct bnx2i_percpu_s *p; struct task_struct *thread; @@ -419,16 +418,17 @@ static void bnx2i_percpu_thread_create(unsigned int cpu) thread = kthread_create_on_node(bnx2i_percpu_io_thread, (void *)p, cpu_to_node(cpu), "bnx2i_thread/%d", cpu); + if (IS_ERR(thread)) + return PTR_ERR(thread); + /* bind thread to the cpu */ - if (likely(!IS_ERR(thread))) { - kthread_bind(thread, cpu); - p->iothread = thread; - wake_up_process(thread); - } + kthread_bind(thread, cpu); + p->iothread = thread; + wake_up_process(thread); + return 0; } - -static void bnx2i_percpu_thread_destroy(unsigned int cpu) +static int bnx2i_cpu_offline(unsigned int cpu) { struct bnx2i_percpu_s *p; struct task_struct *thread; @@ -451,19 +451,6 @@ static void bnx2i_percpu_thread_destroy(unsigned int cpu) spin_unlock_bh(&p->p_work_lock); if (thread) kthread_stop(thread); -} - -static int bnx2i_cpu_online(unsigned int cpu) -{ - pr_info("bnx2i: CPU %x online: Create Rx thread\n", cpu); - bnx2i_percpu_thread_create(cpu); - return 0; -} - -static int bnx2i_cpu_dead(unsigned int cpu) -{ - pr_info("CPU %x offline: Remove Rx thread\n", cpu); - bnx2i_percpu_thread_destroy(cpu); return 0; } @@ -511,28 +498,14 @@ static int __init bnx2i_mod_init(void) p->iothread = NULL; } - get_online_cpus(); - - for_each_online_cpu(cpu) - bnx2i_percpu_thread_create(cpu); - - err = cpuhp_setup_state_nocalls_cpuslocked(CPUHP_AP_ONLINE_DYN, - "scsi/bnx2i:online", - bnx2i_cpu_online, NULL); + err = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "scsi/bnx2i:online", + bnx2i_cpu_online, bnx2i_cpu_offline); if (err < 0) - goto remove_threads; + goto unreg_driver; bnx2i_online_state = err; - - cpuhp_setup_state_nocalls_cpuslocked(CPUHP_SCSI_BNX2I_DEAD, - "scsi/bnx2i:dead", - NULL, bnx2i_cpu_dead); - put_online_cpus(); return 0; -remove_threads: - for_each_online_cpu(cpu) - bnx2i_percpu_thread_destroy(cpu); - put_online_cpus(); +unreg_driver: cnic_unregister_driver(CNIC_ULP_ISCSI); unreg_xport: iscsi_unregister_transport(&bnx2i_iscsi_transport); @@ -552,7 +525,6 @@ out: static void __exit bnx2i_mod_exit(void) { struct bnx2i_hba *hba; - unsigned cpu = 0; mutex_lock(&bnx2i_dev_lock); while (!list_empty(&adapter_list)) { @@ -570,14 +542,7 @@ static void __exit bnx2i_mod_exit(void) } mutex_unlock(&bnx2i_dev_lock); - get_online_cpus(); - - for_each_online_cpu(cpu) - bnx2i_percpu_thread_destroy(cpu); - - cpuhp_remove_state_nocalls_cpuslocked(bnx2i_online_state); - cpuhp_remove_state_nocalls_cpuslocked(CPUHP_SCSI_BNX2I_DEAD); - put_online_cpus(); + cpuhp_remove_state(bnx2i_online_state); iscsi_unregister_transport(&bnx2i_iscsi_transport); cnic_unregister_driver(CNIC_ULP_ISCSI); diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h index 2e7b1731ad12..82b30e638430 100644 --- a/include/linux/cpuhotplug.h +++ b/include/linux/cpuhotplug.h @@ -39,7 +39,6 @@ enum cpuhp_state { CPUHP_PCI_XGENE_DEAD, CPUHP_IOMMU_INTEL_DEAD, CPUHP_LUSTRE_CFS_DEAD, - CPUHP_SCSI_BNX2I_DEAD, CPUHP_WORKQUEUE_PREP, CPUHP_POWER_NUMA_PREPARE, CPUHP_HRTIMERS_PREPARE, -- cgit v1.2.3 From 37ef38f3f83891a2f413fb872bae7d0f9bb95b27 Mon Sep 17 00:00:00 2001 From: Timur Tabi Date: Thu, 27 Jul 2017 16:15:52 -0500 Subject: tty: pl011: fix initialization order of QDF2400 E44 The work-around for Qualcomm Technologies QDF2400 Erratum 44 hinges on a global variable defined in the pl011 driver. The ACPI SPCR parsing code determines whether the work-around is needed, and if so, it changes the console name from "pl011" to "qdf2400_e44". The expectation is that the pl011 driver will implement the work-around when it sees the console name. The global variable qdf2400_e44_present is set when that happens. The problem is that work-around needs to be enabled when the pl011 driver probes, not when the console name is queried. However, sbsa_probe() is called before pl011_console_match(). The work-around appeared to work previously because the default console on QDF2400 platforms was always ttyAMA1. The first time sbsa_probe() is called (for ttyAMA0), qdf2400_e44_present is still false. Then pl011_console_match() is called, and it sets qdf2400_e44_present to true. All subsequent calls to sbsa_probe() enable the work-around. The solution is to move the global variable into spcr.c and let the pl011 driver query it during probe time. This works because all QDF2400 platforms require SPCR, so parse_spcr() will always be called. pl011_console_match still checks for the "qdf2400_e44" console name, but it doesn't do anything else special. Fixes: 5a0722b898f8 ("tty: pl011: use "qdf2400_e44" as the earlycon name for QDF2400 E44") Tested-by: Jeffrey Hugo Signed-off-by: Timur Tabi Signed-off-by: Greg Kroah-Hartman --- drivers/acpi/spcr.c | 36 ++++++++++++++++++++++++++++++++++-- drivers/tty/serial/amba-pl011.c | 37 +++++++++++++++++++------------------ include/linux/acpi.h | 1 + 3 files changed, 54 insertions(+), 20 deletions(-) (limited to 'include/linux') diff --git a/drivers/acpi/spcr.c b/drivers/acpi/spcr.c index 4ac3e06b41d8..98aa8c808a33 100644 --- a/drivers/acpi/spcr.c +++ b/drivers/acpi/spcr.c @@ -16,6 +16,16 @@ #include #include +/* + * Erratum 44 for QDF2432v1 and QDF2400v1 SoCs describes the BUSY bit as + * occasionally getting stuck as 1. To avoid the potential for a hang, check + * TXFE == 0 instead of BUSY == 1. This may not be suitable for all UART + * implementations, so only do so if an affected platform is detected in + * parse_spcr(). + */ +bool qdf2400_e44_present; +EXPORT_SYMBOL(qdf2400_e44_present); + /* * Some Qualcomm Datacenter Technologies SoCs have a defective UART BUSY bit. * Detect them by examining the OEM fields in the SPCR header, similiar to PCI @@ -147,8 +157,30 @@ int __init parse_spcr(bool earlycon) goto done; } - if (qdf2400_erratum_44_present(&table->header)) - uart = "qdf2400_e44"; + /* + * If the E44 erratum is required, then we need to tell the pl011 + * driver to implement the work-around. + * + * The global variable is used by the probe function when it + * creates the UARTs, whether or not they're used as a console. + * + * If the user specifies "traditional" earlycon, the qdf2400_e44 + * console name matches the EARLYCON_DECLARE() statement, and + * SPCR is not used. Parameter "earlycon" is false. + * + * If the user specifies "SPCR" earlycon, then we need to update + * the console name so that it also says "qdf2400_e44". Parameter + * "earlycon" is true. + * + * For consistency, if we change the console name, then we do it + * for everyone, not just earlycon. + */ + if (qdf2400_erratum_44_present(&table->header)) { + qdf2400_e44_present = true; + if (earlycon) + uart = "qdf2400_e44"; + } + if (xgene_8250_erratum_present(table)) iotype = "mmio32"; diff --git a/drivers/tty/serial/amba-pl011.c b/drivers/tty/serial/amba-pl011.c index 8a857bb34fbb..1888d168a41c 100644 --- a/drivers/tty/serial/amba-pl011.c +++ b/drivers/tty/serial/amba-pl011.c @@ -142,15 +142,7 @@ static struct vendor_data vendor_sbsa = { .fixed_options = true, }; -/* - * Erratum 44 for QDF2432v1 and QDF2400v1 SoCs describes the BUSY bit as - * occasionally getting stuck as 1. To avoid the potential for a hang, check - * TXFE == 0 instead of BUSY == 1. This may not be suitable for all UART - * implementations, so only do so if an affected platform is detected in - * parse_spcr(). - */ -static bool qdf2400_e44_present = false; - +#ifdef CONFIG_ACPI_SPCR_TABLE static struct vendor_data vendor_qdt_qdf2400_e44 = { .reg_offset = pl011_std_offsets, .fr_busy = UART011_FR_TXFE, @@ -165,6 +157,7 @@ static struct vendor_data vendor_qdt_qdf2400_e44 = { .always_enabled = true, .fixed_options = true, }; +#endif static u16 pl011_st_offsets[REG_ARRAY_SIZE] = { [REG_DR] = UART01x_DR, @@ -2375,12 +2368,14 @@ static int __init pl011_console_match(struct console *co, char *name, int idx, resource_size_t addr; int i; - if (strcmp(name, "qdf2400_e44") == 0) { - pr_info_once("UART: Working around QDF2400 SoC erratum 44"); - qdf2400_e44_present = true; - } else if (strcmp(name, "pl011") != 0) { + /* + * Systems affected by the Qualcomm Technologies QDF2400 E44 erratum + * have a distinct console name, so make sure we check for that. + * The actual implementation of the erratum occurs in the probe + * function. + */ + if ((strcmp(name, "qdf2400_e44") != 0) && (strcmp(name, "pl011") != 0)) return -ENODEV; - } if (uart_parse_earlycon(options, &iotype, &addr, &options)) return -ENODEV; @@ -2734,11 +2729,17 @@ static int sbsa_uart_probe(struct platform_device *pdev) } uap->port.irq = ret; - uap->reg_offset = vendor_sbsa.reg_offset; - uap->vendor = qdf2400_e44_present ? - &vendor_qdt_qdf2400_e44 : &vendor_sbsa; +#ifdef CONFIG_ACPI_SPCR_TABLE + if (qdf2400_e44_present) { + dev_info(&pdev->dev, "working around QDF2400 SoC erratum 44\n"); + uap->vendor = &vendor_qdt_qdf2400_e44; + } else +#endif + uap->vendor = &vendor_sbsa; + + uap->reg_offset = uap->vendor->reg_offset; uap->fifosize = 32; - uap->port.iotype = vendor_sbsa.access_32b ? UPIO_MEM32 : UPIO_MEM; + uap->port.iotype = uap->vendor->access_32b ? UPIO_MEM32 : UPIO_MEM; uap->port.ops = &sbsa_uart_pops; uap->fixed_baud = baudrate; diff --git a/include/linux/acpi.h b/include/linux/acpi.h index c749eef1daa1..27b4b6615263 100644 --- a/include/linux/acpi.h +++ b/include/linux/acpi.h @@ -1209,6 +1209,7 @@ static inline bool acpi_has_watchdog(void) { return false; } #endif #ifdef CONFIG_ACPI_SPCR_TABLE +extern bool qdf2400_e44_present; int parse_spcr(bool earlycon); #else static inline int parse_spcr(bool earlycon) { return 0; } -- cgit v1.2.3 From 99f828436788f0155798145853607ca8f0e6de93 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 28 Jul 2017 22:29:51 +0100 Subject: dma-buf/sync_file: Allow multiple sync_files to wrap a single dma-fence Up until recently sync_file were create to export a single dma-fence to userspace, and so we could canabalise a bit insie dma-fence to mark whether or not we had enable polling for the sync_file itself. However, with the advent of syncobj, we do allow userspace to create multiple sync_files for a single dma-fence. (Similarly, that the sw-sync validation framework also started returning multiple sync-files wrapping a single dma-fence for a syncpt also triggering the problem.) This patch reverts my suggestion in commit e24165537312 ("dma-buf/sync_file: only enable fence signalling on poll()") to use a single bit in the shared dma-fence and restores the sync_file->flags for tracking the bits individually. Reported-by: Gustavo Padovan Fixes: f1e8c67123cf ("dma-buf/sw-sync: Use an rbtree to sort fences in the timeline") Fixes: e9083420bbac ("drm: introduce sync objects (v4)") Signed-off-by: Chris Wilson Cc: Sumit Semwal Cc: Sean Paul Cc: Gustavo Padovan Cc: dri-devel@lists.freedesktop.org Cc: # v4.13-rc1+ Signed-off-by: Gustavo Padovan Link: http://patchwork.freedesktop.org/patch/msgid/20170728212951.7818-1-chris@chris-wilson.co.uk (cherry picked from commit db1fc97ca0c0d3fdeeadf314d99a26188438940a) --- drivers/dma-buf/sync_file.c | 5 +++-- include/linux/sync_file.h | 3 ++- 2 files changed, 5 insertions(+), 3 deletions(-) (limited to 'include/linux') diff --git a/drivers/dma-buf/sync_file.c b/drivers/dma-buf/sync_file.c index d7e219d2669d..66fb40d0ebdb 100644 --- a/drivers/dma-buf/sync_file.c +++ b/drivers/dma-buf/sync_file.c @@ -304,7 +304,7 @@ static int sync_file_release(struct inode *inode, struct file *file) { struct sync_file *sync_file = file->private_data; - if (test_bit(POLL_ENABLED, &sync_file->fence->flags)) + if (test_bit(POLL_ENABLED, &sync_file->flags)) dma_fence_remove_callback(sync_file->fence, &sync_file->cb); dma_fence_put(sync_file->fence); kfree(sync_file); @@ -318,7 +318,8 @@ static unsigned int sync_file_poll(struct file *file, poll_table *wait) poll_wait(file, &sync_file->wq, wait); - if (!test_and_set_bit(POLL_ENABLED, &sync_file->fence->flags)) { + if (list_empty(&sync_file->cb.node) && + !test_and_set_bit(POLL_ENABLED, &sync_file->flags)) { if (dma_fence_add_callback(sync_file->fence, &sync_file->cb, fence_check_cb_func) < 0) wake_up_all(&sync_file->wq); diff --git a/include/linux/sync_file.h b/include/linux/sync_file.h index 5726107963b2..0ad87c434ae6 100644 --- a/include/linux/sync_file.h +++ b/include/linux/sync_file.h @@ -43,12 +43,13 @@ struct sync_file { #endif wait_queue_head_t wq; + unsigned long flags; struct dma_fence *fence; struct dma_fence_cb cb; }; -#define POLL_ENABLED DMA_FENCE_FLAG_USER_BITS +#define POLL_ENABLED 0 struct sync_file *sync_file_create(struct dma_fence *fence); struct dma_fence *sync_file_get_fence(int fd); -- cgit v1.2.3 From 9c80034921d1ece5c0e3241ba1645bce32387684 Mon Sep 17 00:00:00 2001 From: Wolfram Sang Date: Sat, 29 Jul 2017 14:11:43 +0200 Subject: i2c: rephrase explanation of I2C_CLASS_DEPRECATED Hopefully making clear that it is not needed for new drivers. Signed-off-by: Wolfram Sang --- include/linux/i2c.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'include/linux') diff --git a/include/linux/i2c.h b/include/linux/i2c.h index 00ca5b86a753..d501d3956f13 100644 --- a/include/linux/i2c.h +++ b/include/linux/i2c.h @@ -689,7 +689,8 @@ i2c_unlock_adapter(struct i2c_adapter *adapter) #define I2C_CLASS_HWMON (1<<0) /* lm_sensors, ... */ #define I2C_CLASS_DDC (1<<3) /* DDC bus on graphics adapters */ #define I2C_CLASS_SPD (1<<7) /* Memory modules */ -#define I2C_CLASS_DEPRECATED (1<<8) /* Warn users that adapter will stop using classes */ +/* Warn users that the adapter doesn't support classes anymore */ +#define I2C_CLASS_DEPRECATED (1<<8) /* Internal numbers to terminate lists */ #define I2C_CLIENT_END 0xfffeU -- cgit v1.2.3 From d9535cb7b7603aeb549c697ecdf92024e4d0a650 Mon Sep 17 00:00:00 2001 From: Grygorii Strashko Date: Fri, 28 Jul 2017 17:30:02 -0500 Subject: ptp: introduce ptp auxiliary worker Many PTP drivers required to perform some asynchronous or periodic work, like periodically handling PHC counter overflow or handle delayed timestamp for RX/TX network packets. In most of the cases, such work is implemented using workqueues. Unfortunately, Kernel workqueues might introduce significant delay in work scheduling under high system load and on -RT, which could cause misbehavior of PTP drivers due to internal counter overflow, for example, and there is no way to tune its execution policy and priority manuallly. Hence, The kthread_worker can be used insted of workqueues, as it create separte named kthread for each worker and its its execution policy and priority can be configured using chrt tool. This prblem was reported for two drivers TI CPSW CPTS and dp83640, so instead of modifying each of these driver it was proposed to add PTP auxiliary worker to the PHC subsystem. The patch adds PTP auxiliary worker in PHC subsystem using kthread_worker and kthread_delayed_work and introduces two new PHC subsystem APIs: - long (*do_aux_work)(struct ptp_clock_info *ptp) callback in ptp_clock_info structure, which driver should assign if it require to perform asynchronous or periodic work. Driver should return the delay of the PTP next auxiliary work scheduling time (>=0) or negative value in case further scheduling is not required. - int ptp_schedule_worker(struct ptp_clock *ptp, unsigned long delay) which allows schedule PTP auxiliary work. The name of kthread_worker thread corresponds PTP PHC device name "ptp%d". Signed-off-by: Grygorii Strashko Signed-off-by: David S. Miller --- drivers/ptp/ptp_clock.c | 42 ++++++++++++++++++++++++++++++++++++++++ drivers/ptp/ptp_private.h | 3 +++ include/linux/ptp_clock_kernel.h | 20 +++++++++++++++++++ 3 files changed, 65 insertions(+) (limited to 'include/linux') diff --git a/drivers/ptp/ptp_clock.c b/drivers/ptp/ptp_clock.c index b77435783ef3..7eacc1c4b3b1 100644 --- a/drivers/ptp/ptp_clock.c +++ b/drivers/ptp/ptp_clock.c @@ -28,6 +28,7 @@ #include #include #include +#include #include "ptp_private.h" @@ -184,6 +185,19 @@ static void delete_ptp_clock(struct posix_clock *pc) kfree(ptp); } +static void ptp_aux_kworker(struct kthread_work *work) +{ + struct ptp_clock *ptp = container_of(work, struct ptp_clock, + aux_work.work); + struct ptp_clock_info *info = ptp->info; + long delay; + + delay = info->do_aux_work(info); + + if (delay >= 0) + kthread_queue_delayed_work(ptp->kworker, &ptp->aux_work, delay); +} + /* public interface */ struct ptp_clock *ptp_clock_register(struct ptp_clock_info *info, @@ -217,6 +231,20 @@ struct ptp_clock *ptp_clock_register(struct ptp_clock_info *info, mutex_init(&ptp->pincfg_mux); init_waitqueue_head(&ptp->tsev_wq); + if (ptp->info->do_aux_work) { + char *worker_name = kasprintf(GFP_KERNEL, "ptp%d", ptp->index); + + kthread_init_delayed_work(&ptp->aux_work, ptp_aux_kworker); + ptp->kworker = kthread_create_worker(0, worker_name ? + worker_name : info->name); + kfree(worker_name); + if (IS_ERR(ptp->kworker)) { + err = PTR_ERR(ptp->kworker); + pr_err("failed to create ptp aux_worker %d\n", err); + goto kworker_err; + } + } + err = ptp_populate_pin_groups(ptp); if (err) goto no_pin_groups; @@ -259,6 +287,9 @@ no_pps: no_device: ptp_cleanup_pin_groups(ptp); no_pin_groups: + if (ptp->kworker) + kthread_destroy_worker(ptp->kworker); +kworker_err: mutex_destroy(&ptp->tsevq_mux); mutex_destroy(&ptp->pincfg_mux); ida_simple_remove(&ptp_clocks_map, index); @@ -274,6 +305,11 @@ int ptp_clock_unregister(struct ptp_clock *ptp) ptp->defunct = 1; wake_up_interruptible(&ptp->tsev_wq); + if (ptp->kworker) { + kthread_cancel_delayed_work_sync(&ptp->aux_work); + kthread_destroy_worker(ptp->kworker); + } + /* Release the clock's resources. */ if (ptp->pps_source) pps_unregister_source(ptp->pps_source); @@ -339,6 +375,12 @@ int ptp_find_pin(struct ptp_clock *ptp, } EXPORT_SYMBOL(ptp_find_pin); +int ptp_schedule_worker(struct ptp_clock *ptp, unsigned long delay) +{ + return kthread_mod_delayed_work(ptp->kworker, &ptp->aux_work, delay); +} +EXPORT_SYMBOL(ptp_schedule_worker); + /* module operations */ static void __exit ptp_exit(void) diff --git a/drivers/ptp/ptp_private.h b/drivers/ptp/ptp_private.h index d95888974d0c..b86f1bfecd6f 100644 --- a/drivers/ptp/ptp_private.h +++ b/drivers/ptp/ptp_private.h @@ -22,6 +22,7 @@ #include #include +#include #include #include #include @@ -56,6 +57,8 @@ struct ptp_clock { struct attribute_group pin_attr_group; /* 1st entry is a pointer to the real group, 2nd is NULL terminator */ const struct attribute_group *pin_attr_groups[2]; + struct kthread_worker *kworker; + struct kthread_delayed_work aux_work; }; /* diff --git a/include/linux/ptp_clock_kernel.h b/include/linux/ptp_clock_kernel.h index a026bfd089db..51349d124ee5 100644 --- a/include/linux/ptp_clock_kernel.h +++ b/include/linux/ptp_clock_kernel.h @@ -99,6 +99,11 @@ struct system_device_crosststamp; * parameter func: the desired function to use. * parameter chan: the function channel index to use. * + * @do_work: Request driver to perform auxiliary (periodic) operations + * Driver should return delay of the next auxiliary work scheduling + * time (>=0) or negative value in case further scheduling + * is not required. + * * Drivers should embed their ptp_clock_info within a private * structure, obtaining a reference to it using container_of(). * @@ -126,6 +131,7 @@ struct ptp_clock_info { struct ptp_clock_request *request, int on); int (*verify)(struct ptp_clock_info *ptp, unsigned int pin, enum ptp_pin_function func, unsigned int chan); + long (*do_aux_work)(struct ptp_clock_info *ptp); }; struct ptp_clock; @@ -211,6 +217,16 @@ extern int ptp_clock_index(struct ptp_clock *ptp); int ptp_find_pin(struct ptp_clock *ptp, enum ptp_pin_function func, unsigned int chan); +/** + * ptp_schedule_worker() - schedule ptp auxiliary work + * + * @ptp: The clock obtained from ptp_clock_register(). + * @delay: number of jiffies to wait before queuing + * See kthread_queue_delayed_work() for more info. + */ + +int ptp_schedule_worker(struct ptp_clock *ptp, unsigned long delay); + #else static inline struct ptp_clock *ptp_clock_register(struct ptp_clock_info *info, struct device *parent) @@ -225,6 +241,10 @@ static inline int ptp_clock_index(struct ptp_clock *ptp) static inline int ptp_find_pin(struct ptp_clock *ptp, enum ptp_pin_function func, unsigned int chan) { return -1; } +static inline int ptp_schedule_worker(struct ptp_clock *ptp, + unsigned long delay) +{ return -EOPNOTSUPP; } + #endif #endif -- cgit v1.2.3 From a477b9cd37aa81a490dfa3265b7ff4f2c5a92463 Mon Sep 17 00:00:00 2001 From: Marc Zyngier Date: Tue, 1 Aug 2017 20:11:02 -0500 Subject: PCI: Add pci_reset_function_locked() The implementation of PCI workarounds may require that the device is reset from its probe function. This implies that the PCI device lock is already held, and makes calling pci_reset_function() impossible (since it will itself try to take that lock). Add pci_reset_function_locked(), which is the equivalent of pci_reset_function(), except that it requires the PCI device lock to be already held by the caller. Tested-by: Ard Biesheuvel Signed-off-by: Marc Zyngier [bhelgaas: folded in fix for conflict with 52354b9d1f46 ("PCI: Remove __pci_dev_reset() and pci_dev_reset()")] Signed-off-by: Bjorn Helgaas Cc: stable@vger.kernel.org # 4.11: 52354b9d1f46: PCI: Remove __pci_dev_reset() and pci_dev_reset() Cc: stable@vger.kernel.org # 4.11 --- drivers/pci/pci.c | 35 +++++++++++++++++++++++++++++++++++ include/linux/pci.h | 1 + 2 files changed, 36 insertions(+) (limited to 'include/linux') diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index af0cc3456dc1..b4b7eab29400 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -4259,6 +4259,41 @@ int pci_reset_function(struct pci_dev *dev) } EXPORT_SYMBOL_GPL(pci_reset_function); +/** + * pci_reset_function_locked - quiesce and reset a PCI device function + * @dev: PCI device to reset + * + * Some devices allow an individual function to be reset without affecting + * other functions in the same device. The PCI device must be responsive + * to PCI config space in order to use this function. + * + * This function does not just reset the PCI portion of a device, but + * clears all the state associated with the device. This function differs + * from __pci_reset_function() in that it saves and restores device state + * over the reset. It also differs from pci_reset_function() in that it + * requires the PCI device lock to be held. + * + * Returns 0 if the device function was successfully reset or negative if the + * device doesn't support resetting a single function. + */ +int pci_reset_function_locked(struct pci_dev *dev) +{ + int rc; + + rc = pci_probe_reset_function(dev); + if (rc) + return rc; + + pci_dev_save_and_disable(dev); + + rc = __pci_reset_function_locked(dev); + + pci_dev_restore(dev); + + return rc; +} +EXPORT_SYMBOL_GPL(pci_reset_function_locked); + /** * pci_try_reset_function - quiesce and reset a PCI device function * @dev: PCI device to reset diff --git a/include/linux/pci.h b/include/linux/pci.h index 4869e66dd659..a75c13673852 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h @@ -1067,6 +1067,7 @@ void pcie_flr(struct pci_dev *dev); int __pci_reset_function(struct pci_dev *dev); int __pci_reset_function_locked(struct pci_dev *dev); int pci_reset_function(struct pci_dev *dev); +int pci_reset_function_locked(struct pci_dev *dev); int pci_try_reset_function(struct pci_dev *dev); int pci_probe_reset_slot(struct pci_slot *slot); int pci_reset_slot(struct pci_slot *slot); -- cgit v1.2.3 From 6d29231000bbe0fb9e4893a9c68151ffdd3b5469 Mon Sep 17 00:00:00 2001 From: Boris Brezillon Date: Mon, 31 Jul 2017 10:31:27 +0200 Subject: mtd: nand: Declare tBERS, tR and tPROG as u64 to avoid integer overflow All timings in nand_sdr_timings are expressed in picoseconds but some of them may not fit in an u32. Signed-off-by: Boris Brezillon Fixes: 204e7ecd47e2 ("mtd: nand: Add a few more timings to nand_sdr_timings") Reported-by: Alexander Dahl Cc: Reviewed-by: Alexander Dahl Tested-by: Alexander Dahl Signed-off-by: Boris Brezillon --- drivers/mtd/nand/nand_timings.c | 6 +++--- include/linux/mtd/nand.h | 6 +++--- 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'include/linux') diff --git a/drivers/mtd/nand/nand_timings.c b/drivers/mtd/nand/nand_timings.c index f06312df3669..7e36d7d13c26 100644 --- a/drivers/mtd/nand/nand_timings.c +++ b/drivers/mtd/nand/nand_timings.c @@ -311,9 +311,9 @@ int onfi_init_data_interface(struct nand_chip *chip, struct nand_sdr_timings *timings = &iface->timings.sdr; /* microseconds -> picoseconds */ - timings->tPROG_max = 1000000UL * le16_to_cpu(params->t_prog); - timings->tBERS_max = 1000000UL * le16_to_cpu(params->t_bers); - timings->tR_max = 1000000UL * le16_to_cpu(params->t_r); + timings->tPROG_max = 1000000ULL * le16_to_cpu(params->t_prog); + timings->tBERS_max = 1000000ULL * le16_to_cpu(params->t_bers); + timings->tR_max = 1000000ULL * le16_to_cpu(params->t_r); /* nanoseconds -> picoseconds */ timings->tCCS_min = 1000UL * le16_to_cpu(params->t_ccs); diff --git a/include/linux/mtd/nand.h b/include/linux/mtd/nand.h index 892148c448cc..5216d2eb2289 100644 --- a/include/linux/mtd/nand.h +++ b/include/linux/mtd/nand.h @@ -681,10 +681,10 @@ struct nand_buffers { * @tWW_min: WP# transition to WE# low */ struct nand_sdr_timings { - u32 tBERS_max; + u64 tBERS_max; u32 tCCS_min; - u32 tPROG_max; - u32 tR_max; + u64 tPROG_max; + u64 tR_max; u32 tALH_min; u32 tADL_min; u32 tALS_min; -- cgit v1.2.3 From c994f778bb1cca8ebe7a4e528cefec233e93b5cc Mon Sep 17 00:00:00 2001 From: Inbar Karmy Date: Tue, 1 Aug 2017 16:43:43 +0300 Subject: net/mlx4_en: Fix wrong indication of Wake-on-LAN (WoL) support Currently when WoL is supported but disabled, ethtool reports: "Supports Wake-on: d". Fix the indication of Wol support, so that the indication remains "g" all the time if the NIC supports WoL. Tested: As accepted, when NIC supports WoL- ethtool reports: Supports Wake-on: g Wake-on: d when NIC doesn't support WoL- ethtool reports: Supports Wake-on: d Wake-on: d Fixes: 14c07b1358ed ("mlx4: Wake on LAN support") Signed-off-by: Inbar Karmy Signed-off-by: Tariq Toukan Signed-off-by: David S. Miller --- drivers/net/ethernet/mellanox/mlx4/en_ethtool.c | 15 ++++++++------- drivers/net/ethernet/mellanox/mlx4/fw.c | 4 ++++ drivers/net/ethernet/mellanox/mlx4/fw.h | 1 + drivers/net/ethernet/mellanox/mlx4/main.c | 2 ++ include/linux/mlx4/device.h | 1 + 5 files changed, 16 insertions(+), 7 deletions(-) (limited to 'include/linux') diff --git a/drivers/net/ethernet/mellanox/mlx4/en_ethtool.c b/drivers/net/ethernet/mellanox/mlx4/en_ethtool.c index c751a1d434ad..3d4e4a5d00d1 100644 --- a/drivers/net/ethernet/mellanox/mlx4/en_ethtool.c +++ b/drivers/net/ethernet/mellanox/mlx4/en_ethtool.c @@ -223,6 +223,7 @@ static void mlx4_en_get_wol(struct net_device *netdev, struct ethtool_wolinfo *wol) { struct mlx4_en_priv *priv = netdev_priv(netdev); + struct mlx4_caps *caps = &priv->mdev->dev->caps; int err = 0; u64 config = 0; u64 mask; @@ -235,24 +236,24 @@ static void mlx4_en_get_wol(struct net_device *netdev, mask = (priv->port == 1) ? MLX4_DEV_CAP_FLAG_WOL_PORT1 : MLX4_DEV_CAP_FLAG_WOL_PORT2; - if (!(priv->mdev->dev->caps.flags & mask)) { + if (!(caps->flags & mask)) { wol->supported = 0; wol->wolopts = 0; return; } + if (caps->wol_port[priv->port]) + wol->supported = WAKE_MAGIC; + else + wol->supported = 0; + err = mlx4_wol_read(priv->mdev->dev, &config, priv->port); if (err) { en_err(priv, "Failed to get WoL information\n"); return; } - if (config & MLX4_EN_WOL_MAGIC) - wol->supported = WAKE_MAGIC; - else - wol->supported = 0; - - if (config & MLX4_EN_WOL_ENABLED) + if ((config & MLX4_EN_WOL_ENABLED) && (config & MLX4_EN_WOL_MAGIC)) wol->wolopts = WAKE_MAGIC; else wol->wolopts = 0; diff --git a/drivers/net/ethernet/mellanox/mlx4/fw.c b/drivers/net/ethernet/mellanox/mlx4/fw.c index 37e84a59e751..c165f16623a9 100644 --- a/drivers/net/ethernet/mellanox/mlx4/fw.c +++ b/drivers/net/ethernet/mellanox/mlx4/fw.c @@ -764,6 +764,7 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap) #define QUERY_DEV_CAP_CQ_TS_SUPPORT_OFFSET 0x3e #define QUERY_DEV_CAP_MAX_PKEY_OFFSET 0x3f #define QUERY_DEV_CAP_EXT_FLAGS_OFFSET 0x40 +#define QUERY_DEV_CAP_WOL_OFFSET 0x43 #define QUERY_DEV_CAP_FLAGS_OFFSET 0x44 #define QUERY_DEV_CAP_RSVD_UAR_OFFSET 0x48 #define QUERY_DEV_CAP_UAR_SZ_OFFSET 0x49 @@ -920,6 +921,9 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap) MLX4_GET(ext_flags, outbox, QUERY_DEV_CAP_EXT_FLAGS_OFFSET); MLX4_GET(flags, outbox, QUERY_DEV_CAP_FLAGS_OFFSET); dev_cap->flags = flags | (u64)ext_flags << 32; + MLX4_GET(field, outbox, QUERY_DEV_CAP_WOL_OFFSET); + dev_cap->wol_port[1] = !!(field & 0x20); + dev_cap->wol_port[2] = !!(field & 0x40); MLX4_GET(field, outbox, QUERY_DEV_CAP_RSVD_UAR_OFFSET); dev_cap->reserved_uars = field >> 4; MLX4_GET(field, outbox, QUERY_DEV_CAP_UAR_SZ_OFFSET); diff --git a/drivers/net/ethernet/mellanox/mlx4/fw.h b/drivers/net/ethernet/mellanox/mlx4/fw.h index 5343a0599253..b52ba01aa486 100644 --- a/drivers/net/ethernet/mellanox/mlx4/fw.h +++ b/drivers/net/ethernet/mellanox/mlx4/fw.h @@ -129,6 +129,7 @@ struct mlx4_dev_cap { u32 dmfs_high_rate_qpn_range; struct mlx4_rate_limit_caps rl_caps; struct mlx4_port_cap port_cap[MLX4_MAX_PORTS + 1]; + bool wol_port[MLX4_MAX_PORTS + 1]; }; struct mlx4_func_cap { diff --git a/drivers/net/ethernet/mellanox/mlx4/main.c b/drivers/net/ethernet/mellanox/mlx4/main.c index a27c9c13a36e..09b9bc17bce9 100644 --- a/drivers/net/ethernet/mellanox/mlx4/main.c +++ b/drivers/net/ethernet/mellanox/mlx4/main.c @@ -424,6 +424,8 @@ static int mlx4_dev_cap(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap) dev->caps.stat_rate_support = dev_cap->stat_rate_support; dev->caps.max_gso_sz = dev_cap->max_gso_sz; dev->caps.max_rss_tbl_sz = dev_cap->max_rss_tbl_sz; + dev->caps.wol_port[1] = dev_cap->wol_port[1]; + dev->caps.wol_port[2] = dev_cap->wol_port[2]; /* Save uar page shift */ if (!mlx4_is_slave(dev)) { diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h index aad5d81dfb44..b54517c05e9a 100644 --- a/include/linux/mlx4/device.h +++ b/include/linux/mlx4/device.h @@ -620,6 +620,7 @@ struct mlx4_caps { u32 dmfs_high_rate_qpn_base; u32 dmfs_high_rate_qpn_range; u32 vf_caps; + bool wol_port[MLX4_MAX_PORTS + 1]; struct mlx4_rate_limit_caps rl_caps; }; -- cgit v1.2.3 From 931b3c1a832621b4bdcbaf783096fc267eb36fbe Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Tue, 1 Aug 2017 09:41:37 +0300 Subject: RDMA/mlx5: Fix existence check for extended address vector The extended address vector is the highest bit in be32 variable, but it was compared with the lowest. This patch fixes the endianness of that check and removes already declared define. Fixes: 17d2f88f92ce ("IB/mlx5: Add ODP atomics support") Reviewed-by: Artemy Kovalyov Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/odp.c | 2 +- include/linux/mlx5/qp.h | 1 - 2 files changed, 1 insertion(+), 2 deletions(-) (limited to 'include/linux') diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c index ae0746754008..3d701c7a4c91 100644 --- a/drivers/infiniband/hw/mlx5/odp.c +++ b/drivers/infiniband/hw/mlx5/odp.c @@ -939,7 +939,7 @@ static int mlx5_ib_mr_initiator_pfault_handler( if (qp->ibqp.qp_type != IB_QPT_RC) { av = *wqe; - if (av->dqp_dct & be32_to_cpu(MLX5_WQE_AV_EXT)) + if (av->dqp_dct & cpu_to_be32(MLX5_EXTENDED_UD_AV)) *wqe += sizeof(struct mlx5_av); else *wqe += sizeof(struct mlx5_base_av); diff --git a/include/linux/mlx5/qp.h b/include/linux/mlx5/qp.h index 6f41270d80c0..f378dc0e7eaf 100644 --- a/include/linux/mlx5/qp.h +++ b/include/linux/mlx5/qp.h @@ -212,7 +212,6 @@ struct mlx5_wqe_ctrl_seg { #define MLX5_WQE_CTRL_OPCODE_MASK 0xff #define MLX5_WQE_CTRL_WQE_INDEX_MASK 0x00ffff00 #define MLX5_WQE_CTRL_WQE_INDEX_SHIFT 8 -#define MLX5_WQE_AV_EXT 0x80000000 enum { MLX5_ETH_WQE_L3_INNER_CSUM = 1 << 4, -- cgit v1.2.3 From 0cca6c8920ade95e2741b2062cf1397dc546fb0f Mon Sep 17 00:00:00 2001 From: Ludovic Desroches Date: Sun, 6 Aug 2017 16:00:05 +0200 Subject: pinctrl: generic: update references to Documentation/pinctrl.txt Update deprecated references to Documentation/pinctrl.txt since it has been moved to Documentation/driver-api/pinctl.rst. Signed-off-by: Ludovic Desroches Fixes: 5a9b73832e9e ("pinctrl.txt: move it to the driver-api book") Signed-off-by: Linus Walleij --- Documentation/gpio/gpio-legacy.txt | 2 +- MAINTAINERS | 2 +- include/linux/device.h | 2 +- include/linux/pinctrl/pinconf-generic.h | 4 ++-- 4 files changed, 5 insertions(+), 5 deletions(-) (limited to 'include/linux') diff --git a/Documentation/gpio/gpio-legacy.txt b/Documentation/gpio/gpio-legacy.txt index b34fd94f7089..5eacc147ea87 100644 --- a/Documentation/gpio/gpio-legacy.txt +++ b/Documentation/gpio/gpio-legacy.txt @@ -459,7 +459,7 @@ pin controller? This is done by registering "ranges" of pins, which are essentially cross-reference tables. These are described in -Documentation/pinctrl.txt +Documentation/driver-api/pinctl.rst While the pin allocation is totally managed by the pinctrl subsystem, gpio (under gpiolib) is still maintained by gpio drivers. It may happen diff --git a/MAINTAINERS b/MAINTAINERS index f66488dfdbc9..2c85558aec19 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -10375,7 +10375,7 @@ L: linux-gpio@vger.kernel.org T: git git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl.git S: Maintained F: Documentation/devicetree/bindings/pinctrl/ -F: Documentation/pinctrl.txt +F: Documentation/driver-api/pinctl.rst F: drivers/pinctrl/ F: include/linux/pinctrl/ diff --git a/include/linux/device.h b/include/linux/device.h index 723cd54b94da..beabdbc08420 100644 --- a/include/linux/device.h +++ b/include/linux/device.h @@ -843,7 +843,7 @@ struct dev_links_info { * hibernation, system resume and during runtime PM transitions * along with subsystem-level and driver-level callbacks. * @pins: For device pin management. - * See Documentation/pinctrl.txt for details. + * See Documentation/driver-api/pinctl.rst for details. * @msi_list: Hosts MSI descriptors * @msi_domain: The generic MSI domain this device is using. * @numa_node: NUMA node this device is close to. diff --git a/include/linux/pinctrl/pinconf-generic.h b/include/linux/pinctrl/pinconf-generic.h index 231d3075815a..e91d1b6a260d 100644 --- a/include/linux/pinctrl/pinconf-generic.h +++ b/include/linux/pinctrl/pinconf-generic.h @@ -81,8 +81,8 @@ * it. * @PIN_CONFIG_OUTPUT: this will configure the pin as an output and drive a * value on the line. Use argument 1 to indicate high level, argument 0 to - * indicate low level. (Please see Documentation/pinctrl.txt, section - * "GPIO mode pitfalls" for a discussion around this parameter.) + * indicate low level. (Please see Documentation/driver-api/pinctl.rst, + * section "GPIO mode pitfalls" for a discussion around this parameter.) * @PIN_CONFIG_POWER_SOURCE: if the pin can select between different power * supplies, the argument to this parameter (on a custom format) tells * the driver which alternative power source to use. -- cgit v1.2.3 From 0fb228d30b8d72bfee51f57e638d412324d44a11 Mon Sep 17 00:00:00 2001 From: James Smart Date: Tue, 1 Aug 2017 15:12:39 -0700 Subject: nvmet_fc: add defer_req callback for deferment of cmd buffer return At queue creation, the transport allocates a local job struct (struct nvmet_fc_fcp_iod) for each possible element of the queue. When a new CMD is received from the wire, a jobs struct is allocated from the queue and then used for the duration of the command. The job struct contains buffer space for the wire command iu. Thus, upon allocation of the job struct, the cmd iu buffer is copied to the job struct and the LLDD may immediately free/reuse the CMD IU buffer passed in the call. However, in some circumstances, due to the packetized nature of FC and the api of the FC LLDD which may issue a hw command to send the wire response, but the LLDD may not get the hw completion for the command and upcall the nvmet_fc layer before a new command may be asynchronously received on the wire. In other words, its possible for the initiator to get the response from the wire, thus believe a command slot free, and send a new command iu. The new command iu may be received by the LLDD and passed to the transport before the LLDD had serviced the hw completion and made the teardown calls for the original job struct. As such, there is no available job struct available for the new io. E.g. it appears like the host sent more queue elements than the queue size. It didn't based on it's understanding. Rather than treat this as a hard connection failure queue the new request until the job struct does free up. As the buffer isn't copied as there's no job struct, a special return value must be returned to the LLDD to signify to hold off on recycling the cmd iu buffer. And later, when a job struct is allocated and the buffer copied, a new LLDD callback is introduced to notify the LLDD and allow it to recycle it's command iu buffer. Signed-off-by: James Smart Reviewed-by: Johannes Thumshirn Signed-off-by: Christoph Hellwig --- drivers/nvme/target/fc.c | 212 +++++++++++++++++++++++++++++++++++------ include/linux/nvme-fc-driver.h | 7 ++ 2 files changed, 191 insertions(+), 28 deletions(-) (limited to 'include/linux') diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c index 31ca55dfcb1d..1b7f2520a20d 100644 --- a/drivers/nvme/target/fc.c +++ b/drivers/nvme/target/fc.c @@ -114,6 +114,11 @@ struct nvmet_fc_tgtport { struct kref ref; }; +struct nvmet_fc_defer_fcp_req { + struct list_head req_list; + struct nvmefc_tgt_fcp_req *fcp_req; +}; + struct nvmet_fc_tgt_queue { bool ninetypercent; u16 qid; @@ -132,6 +137,8 @@ struct nvmet_fc_tgt_queue { struct nvmet_fc_tgt_assoc *assoc; struct nvmet_fc_fcp_iod *fod; /* array of fcp_iods */ struct list_head fod_list; + struct list_head pending_cmd_list; + struct list_head avail_defer_list; struct workqueue_struct *work_q; struct kref ref; } __aligned(sizeof(unsigned long long)); @@ -223,6 +230,8 @@ static void nvmet_fc_tgt_q_put(struct nvmet_fc_tgt_queue *queue); static int nvmet_fc_tgt_q_get(struct nvmet_fc_tgt_queue *queue); static void nvmet_fc_tgtport_put(struct nvmet_fc_tgtport *tgtport); static int nvmet_fc_tgtport_get(struct nvmet_fc_tgtport *tgtport); +static void nvmet_fc_handle_fcp_rqst(struct nvmet_fc_tgtport *tgtport, + struct nvmet_fc_fcp_iod *fod); /* *********************** FC-NVME DMA Handling **************************** */ @@ -463,9 +472,9 @@ static struct nvmet_fc_fcp_iod * nvmet_fc_alloc_fcp_iod(struct nvmet_fc_tgt_queue *queue) { static struct nvmet_fc_fcp_iod *fod; - unsigned long flags; - spin_lock_irqsave(&queue->qlock, flags); + lockdep_assert_held(&queue->qlock); + fod = list_first_entry_or_null(&queue->fod_list, struct nvmet_fc_fcp_iod, fcp_list); if (fod) { @@ -477,17 +486,37 @@ nvmet_fc_alloc_fcp_iod(struct nvmet_fc_tgt_queue *queue) * will "inherit" that reference. */ } - spin_unlock_irqrestore(&queue->qlock, flags); return fod; } +static void +nvmet_fc_queue_fcp_req(struct nvmet_fc_tgtport *tgtport, + struct nvmet_fc_tgt_queue *queue, + struct nvmefc_tgt_fcp_req *fcpreq) +{ + struct nvmet_fc_fcp_iod *fod = fcpreq->nvmet_fc_private; + + /* + * put all admin cmds on hw queue id 0. All io commands go to + * the respective hw queue based on a modulo basis + */ + fcpreq->hwqid = queue->qid ? + ((queue->qid - 1) % tgtport->ops->max_hw_queues) : 0; + + if (tgtport->ops->target_features & NVMET_FCTGTFEAT_CMD_IN_ISR) + queue_work_on(queue->cpu, queue->work_q, &fod->work); + else + nvmet_fc_handle_fcp_rqst(tgtport, fod); +} + static void nvmet_fc_free_fcp_iod(struct nvmet_fc_tgt_queue *queue, struct nvmet_fc_fcp_iod *fod) { struct nvmefc_tgt_fcp_req *fcpreq = fod->fcpreq; struct nvmet_fc_tgtport *tgtport = fod->tgtport; + struct nvmet_fc_defer_fcp_req *deferfcp; unsigned long flags; fc_dma_sync_single_for_cpu(tgtport->dev, fod->rspdma, @@ -495,21 +524,56 @@ nvmet_fc_free_fcp_iod(struct nvmet_fc_tgt_queue *queue, fcpreq->nvmet_fc_private = NULL; - spin_lock_irqsave(&queue->qlock, flags); - list_add_tail(&fod->fcp_list, &fod->queue->fod_list); fod->active = false; fod->abort = false; fod->aborted = false; fod->writedataactive = false; fod->fcpreq = NULL; + + tgtport->ops->fcp_req_release(&tgtport->fc_target_port, fcpreq); + + spin_lock_irqsave(&queue->qlock, flags); + deferfcp = list_first_entry_or_null(&queue->pending_cmd_list, + struct nvmet_fc_defer_fcp_req, req_list); + if (!deferfcp) { + list_add_tail(&fod->fcp_list, &fod->queue->fod_list); + spin_unlock_irqrestore(&queue->qlock, flags); + + /* Release reference taken at queue lookup and fod allocation */ + nvmet_fc_tgt_q_put(queue); + return; + } + + /* Re-use the fod for the next pending cmd that was deferred */ + list_del(&deferfcp->req_list); + + fcpreq = deferfcp->fcp_req; + + /* deferfcp can be reused for another IO at a later date */ + list_add_tail(&deferfcp->req_list, &queue->avail_defer_list); + spin_unlock_irqrestore(&queue->qlock, flags); + /* Save NVME CMD IO in fod */ + memcpy(&fod->cmdiubuf, fcpreq->rspaddr, fcpreq->rsplen); + + /* Setup new fcpreq to be processed */ + fcpreq->rspaddr = NULL; + fcpreq->rsplen = 0; + fcpreq->nvmet_fc_private = fod; + fod->fcpreq = fcpreq; + fod->active = true; + + /* inform LLDD IO is now being processed */ + tgtport->ops->defer_rcv(&tgtport->fc_target_port, fcpreq); + + /* Submit deferred IO for processing */ + nvmet_fc_queue_fcp_req(tgtport, queue, fcpreq); + /* - * release the reference taken at queue lookup and fod allocation + * Leave the queue lookup get reference taken when + * fod was originally allocated. */ - nvmet_fc_tgt_q_put(queue); - - tgtport->ops->fcp_req_release(&tgtport->fc_target_port, fcpreq); } static int @@ -569,6 +633,8 @@ nvmet_fc_alloc_target_queue(struct nvmet_fc_tgt_assoc *assoc, queue->port = assoc->tgtport->port; queue->cpu = nvmet_fc_queue_to_cpu(assoc->tgtport, qid); INIT_LIST_HEAD(&queue->fod_list); + INIT_LIST_HEAD(&queue->avail_defer_list); + INIT_LIST_HEAD(&queue->pending_cmd_list); atomic_set(&queue->connected, 0); atomic_set(&queue->sqtail, 0); atomic_set(&queue->rsn, 1); @@ -638,6 +704,7 @@ nvmet_fc_delete_target_queue(struct nvmet_fc_tgt_queue *queue) { struct nvmet_fc_tgtport *tgtport = queue->assoc->tgtport; struct nvmet_fc_fcp_iod *fod = queue->fod; + struct nvmet_fc_defer_fcp_req *deferfcp; unsigned long flags; int i, writedataactive; bool disconnect; @@ -666,6 +733,35 @@ nvmet_fc_delete_target_queue(struct nvmet_fc_tgt_queue *queue) } } } + + /* Cleanup defer'ed IOs in queue */ + list_for_each_entry(deferfcp, &queue->avail_defer_list, req_list) { + list_del(&deferfcp->req_list); + kfree(deferfcp); + } + + for (;;) { + deferfcp = list_first_entry_or_null(&queue->pending_cmd_list, + struct nvmet_fc_defer_fcp_req, req_list); + if (!deferfcp) + break; + + list_del(&deferfcp->req_list); + spin_unlock_irqrestore(&queue->qlock, flags); + + tgtport->ops->defer_rcv(&tgtport->fc_target_port, + deferfcp->fcp_req); + + tgtport->ops->fcp_abort(&tgtport->fc_target_port, + deferfcp->fcp_req); + + tgtport->ops->fcp_req_release(&tgtport->fc_target_port, + deferfcp->fcp_req); + + kfree(deferfcp); + + spin_lock_irqsave(&queue->qlock, flags); + } spin_unlock_irqrestore(&queue->qlock, flags); flush_workqueue(queue->work_q); @@ -2172,11 +2268,38 @@ nvmet_fc_handle_fcp_rqst_work(struct work_struct *work) * Pass a FC-NVME FCP CMD IU received from the FC link to the nvmet-fc * layer for processing. * - * The nvmet-fc layer will copy cmd payload to an internal structure for - * processing. As such, upon completion of the routine, the LLDD may - * immediately free/reuse the CMD IU buffer passed in the call. + * The nvmet_fc layer allocates a local job structure (struct + * nvmet_fc_fcp_iod) from the queue for the io and copies the + * CMD IU buffer to the job structure. As such, on a successful + * completion (returns 0), the LLDD may immediately free/reuse + * the CMD IU buffer passed in the call. + * + * However, in some circumstances, due to the packetized nature of FC + * and the api of the FC LLDD which may issue a hw command to send the + * response, but the LLDD may not get the hw completion for that command + * and upcall the nvmet_fc layer before a new command may be + * asynchronously received - its possible for a command to be received + * before the LLDD and nvmet_fc have recycled the job structure. It gives + * the appearance of more commands received than fits in the sq. + * To alleviate this scenario, a temporary queue is maintained in the + * transport for pending LLDD requests waiting for a queue job structure. + * In these "overrun" cases, a temporary queue element is allocated + * the LLDD request and CMD iu buffer information remembered, and the + * routine returns a -EOVERFLOW status. Subsequently, when a queue job + * structure is freed, it is immediately reallocated for anything on the + * pending request list. The LLDDs defer_rcv() callback is called, + * informing the LLDD that it may reuse the CMD IU buffer, and the io + * is then started normally with the transport. * - * If this routine returns error, the lldd should abort the exchange. + * The LLDD, when receiving an -EOVERFLOW completion status, is to treat + * the completion as successful but must not reuse the CMD IU buffer + * until the LLDD's defer_rcv() callback has been called for the + * corresponding struct nvmefc_tgt_fcp_req pointer. + * + * If there is any other condition in which an error occurs, the + * transport will return a non-zero status indicating the error. + * In all cases other than -EOVERFLOW, the transport has not accepted the + * request and the LLDD should abort the exchange. * * @target_port: pointer to the (registered) target port the FCP CMD IU * was received on. @@ -2194,6 +2317,8 @@ nvmet_fc_rcv_fcp_req(struct nvmet_fc_target_port *target_port, struct nvme_fc_cmd_iu *cmdiu = cmdiubuf; struct nvmet_fc_tgt_queue *queue; struct nvmet_fc_fcp_iod *fod; + struct nvmet_fc_defer_fcp_req *deferfcp; + unsigned long flags; /* validate iu, so the connection id can be used to find the queue */ if ((cmdiubuf_len != sizeof(*cmdiu)) || @@ -2214,29 +2339,60 @@ nvmet_fc_rcv_fcp_req(struct nvmet_fc_target_port *target_port, * when the fod is freed. */ + spin_lock_irqsave(&queue->qlock, flags); + fod = nvmet_fc_alloc_fcp_iod(queue); - if (!fod) { + if (fod) { + spin_unlock_irqrestore(&queue->qlock, flags); + + fcpreq->nvmet_fc_private = fod; + fod->fcpreq = fcpreq; + + memcpy(&fod->cmdiubuf, cmdiubuf, cmdiubuf_len); + + nvmet_fc_queue_fcp_req(tgtport, queue, fcpreq); + + return 0; + } + + if (!tgtport->ops->defer_rcv) { + spin_unlock_irqrestore(&queue->qlock, flags); /* release the queue lookup reference */ nvmet_fc_tgt_q_put(queue); return -ENOENT; } - fcpreq->nvmet_fc_private = fod; - fod->fcpreq = fcpreq; - /* - * put all admin cmds on hw queue id 0. All io commands go to - * the respective hw queue based on a modulo basis - */ - fcpreq->hwqid = queue->qid ? - ((queue->qid - 1) % tgtport->ops->max_hw_queues) : 0; - memcpy(&fod->cmdiubuf, cmdiubuf, cmdiubuf_len); + deferfcp = list_first_entry_or_null(&queue->avail_defer_list, + struct nvmet_fc_defer_fcp_req, req_list); + if (deferfcp) { + /* Just re-use one that was previously allocated */ + list_del(&deferfcp->req_list); + } else { + spin_unlock_irqrestore(&queue->qlock, flags); - if (tgtport->ops->target_features & NVMET_FCTGTFEAT_CMD_IN_ISR) - queue_work_on(queue->cpu, queue->work_q, &fod->work); - else - nvmet_fc_handle_fcp_rqst(tgtport, fod); + /* Now we need to dynamically allocate one */ + deferfcp = kmalloc(sizeof(*deferfcp), GFP_KERNEL); + if (!deferfcp) { + /* release the queue lookup reference */ + nvmet_fc_tgt_q_put(queue); + return -ENOMEM; + } + spin_lock_irqsave(&queue->qlock, flags); + } - return 0; + /* For now, use rspaddr / rsplen to save payload information */ + fcpreq->rspaddr = cmdiubuf; + fcpreq->rsplen = cmdiubuf_len; + deferfcp->fcp_req = fcpreq; + + /* defer processing till a fod becomes available */ + list_add_tail(&deferfcp->req_list, &queue->pending_cmd_list); + + /* NOTE: the queue lookup reference is still valid */ + + spin_unlock_irqrestore(&queue->qlock, flags); + + return -EOVERFLOW; } EXPORT_SYMBOL_GPL(nvmet_fc_rcv_fcp_req); diff --git a/include/linux/nvme-fc-driver.h b/include/linux/nvme-fc-driver.h index 6c8c5d8041b7..2591878c1d48 100644 --- a/include/linux/nvme-fc-driver.h +++ b/include/linux/nvme-fc-driver.h @@ -346,6 +346,11 @@ struct nvme_fc_remote_port { * indicating an FC transport Aborted status. * Entrypoint is Mandatory. * + * @defer_rcv: Called by the transport to signal the LLLD that it has + * begun processing of a previously received NVME CMD IU. The LLDD + * is now free to re-use the rcv buffer associated with the + * nvmefc_tgt_fcp_req. + * * @max_hw_queues: indicates the maximum number of hw queues the LLDD * supports for cpu affinitization. * Value is Mandatory. Must be at least 1. @@ -846,6 +851,8 @@ struct nvmet_fc_target_template { struct nvmefc_tgt_fcp_req *fcpreq); void (*fcp_req_release)(struct nvmet_fc_target_port *tgtport, struct nvmefc_tgt_fcp_req *fcpreq); + void (*defer_rcv)(struct nvmet_fc_target_port *tgtport, + struct nvmefc_tgt_fcp_req *fcpreq); u32 max_hw_queues; u16 max_sgl_segments; -- cgit v1.2.3 From 16af97dc5a8975371a83d9e30a64038b48f40a2d Mon Sep 17 00:00:00 2001 From: Nadav Amit Date: Thu, 10 Aug 2017 15:23:56 -0700 Subject: mm: migrate: prevent racy access to tlb_flush_pending Patch series "fixes of TLB batching races", v6. It turns out that Linux TLB batching mechanism suffers from various races. Races that are caused due to batching during reclamation were recently handled by Mel and this patch-set deals with others. The more fundamental issue is that concurrent updates of the page-tables allow for TLB flushes to be batched on one core, while another core changes the page-tables. This other core may assume a PTE change does not require a flush based on the updated PTE value, while it is unaware that TLB flushes are still pending. This behavior affects KSM (which may result in memory corruption) and MADV_FREE and MADV_DONTNEED (which may result in incorrect behavior). A proof-of-concept can easily produce the wrong behavior of MADV_DONTNEED. Memory corruption in KSM is harder to produce in practice, but was observed by hacking the kernel and adding a delay before flushing and replacing the KSM page. Finally, there is also one memory barrier missing, which may affect architectures with weak memory model. This patch (of 7): Setting and clearing mm->tlb_flush_pending can be performed by multiple threads, since mmap_sem may only be acquired for read in task_numa_work(). If this happens, tlb_flush_pending might be cleared while one of the threads still changes PTEs and batches TLB flushes. This can lead to the same race between migration and change_protection_range() that led to the introduction of tlb_flush_pending. The result of this race was data corruption, which means that this patch also addresses a theoretically possible data corruption. An actual data corruption was not observed, yet the race was was confirmed by adding assertion to check tlb_flush_pending is not set by two threads, adding artificial latency in change_protection_range() and using sysctl to reduce kernel.numa_balancing_scan_delay_ms. Link: http://lkml.kernel.org/r/20170802000818.4760-2-namit@vmware.com Fixes: 20841405940e ("mm: fix TLB flush race between migration, and change_protection_range") Signed-off-by: Nadav Amit Acked-by: Mel Gorman Acked-by: Rik van Riel Acked-by: Minchan Kim Cc: Andy Lutomirski Cc: Hugh Dickins Cc: "David S. Miller" Cc: Andrea Arcangeli Cc: Heiko Carstens Cc: Ingo Molnar Cc: Jeff Dike Cc: Martin Schwidefsky Cc: Mel Gorman Cc: Russell King Cc: Sergey Senozhatsky Cc: Tony Luck Cc: Yoshinori Sato Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/mm_types.h | 31 ++++++++++++++++++++++--------- kernel/fork.c | 2 +- mm/debug.c | 2 +- mm/mprotect.c | 4 ++-- 4 files changed, 26 insertions(+), 13 deletions(-) (limited to 'include/linux') diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h index 7f384bb62d8e..f58f76ee1dfa 100644 --- a/include/linux/mm_types.h +++ b/include/linux/mm_types.h @@ -493,7 +493,7 @@ struct mm_struct { * can move process memory needs to flush the TLB when moving a * PROT_NONE or PROT_NUMA mapped page. */ - bool tlb_flush_pending; + atomic_t tlb_flush_pending; #endif #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH /* See flush_tlb_batched_pending() */ @@ -532,33 +532,46 @@ static inline cpumask_t *mm_cpumask(struct mm_struct *mm) static inline bool mm_tlb_flush_pending(struct mm_struct *mm) { barrier(); - return mm->tlb_flush_pending; + return atomic_read(&mm->tlb_flush_pending) > 0; } -static inline void set_tlb_flush_pending(struct mm_struct *mm) + +static inline void init_tlb_flush_pending(struct mm_struct *mm) { - mm->tlb_flush_pending = true; + atomic_set(&mm->tlb_flush_pending, 0); +} + +static inline void inc_tlb_flush_pending(struct mm_struct *mm) +{ + atomic_inc(&mm->tlb_flush_pending); /* - * Guarantee that the tlb_flush_pending store does not leak into the + * Guarantee that the tlb_flush_pending increase does not leak into the * critical section updating the page tables */ smp_mb__before_spinlock(); } + /* Clearing is done after a TLB flush, which also provides a barrier. */ -static inline void clear_tlb_flush_pending(struct mm_struct *mm) +static inline void dec_tlb_flush_pending(struct mm_struct *mm) { barrier(); - mm->tlb_flush_pending = false; + atomic_dec(&mm->tlb_flush_pending); } #else static inline bool mm_tlb_flush_pending(struct mm_struct *mm) { return false; } -static inline void set_tlb_flush_pending(struct mm_struct *mm) + +static inline void init_tlb_flush_pending(struct mm_struct *mm) { } -static inline void clear_tlb_flush_pending(struct mm_struct *mm) + +static inline void inc_tlb_flush_pending(struct mm_struct *mm) +{ +} + +static inline void dec_tlb_flush_pending(struct mm_struct *mm) { } #endif diff --git a/kernel/fork.c b/kernel/fork.c index 17921b0390b4..e075b7780421 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -807,7 +807,7 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, mm_init_aio(mm); mm_init_owner(mm, p); mmu_notifier_mm_init(mm); - clear_tlb_flush_pending(mm); + init_tlb_flush_pending(mm); #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS mm->pmd_huge_pte = NULL; #endif diff --git a/mm/debug.c b/mm/debug.c index db1cd26d8752..d70103bb4731 100644 --- a/mm/debug.c +++ b/mm/debug.c @@ -159,7 +159,7 @@ void dump_mm(const struct mm_struct *mm) mm->numa_next_scan, mm->numa_scan_offset, mm->numa_scan_seq, #endif #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) - mm->tlb_flush_pending, + atomic_read(&mm->tlb_flush_pending), #endif mm->def_flags, &mm->def_flags ); diff --git a/mm/mprotect.c b/mm/mprotect.c index 4180ad8cc9c5..bd0f409922cb 100644 --- a/mm/mprotect.c +++ b/mm/mprotect.c @@ -244,7 +244,7 @@ static unsigned long change_protection_range(struct vm_area_struct *vma, BUG_ON(addr >= end); pgd = pgd_offset(mm, addr); flush_cache_range(vma, addr, end); - set_tlb_flush_pending(mm); + inc_tlb_flush_pending(mm); do { next = pgd_addr_end(addr, end); if (pgd_none_or_clear_bad(pgd)) @@ -256,7 +256,7 @@ static unsigned long change_protection_range(struct vm_area_struct *vma, /* Only flush the TLB if we actually modified any entries: */ if (pages) flush_tlb_range(vma, start, end); - clear_tlb_flush_pending(mm); + dec_tlb_flush_pending(mm); return pages; } -- cgit v1.2.3 From 0a2c40487f3e4215c6ab46e7f837036badfb542b Mon Sep 17 00:00:00 2001 From: Nadav Amit Date: Thu, 10 Aug 2017 15:23:59 -0700 Subject: mm: migrate: fix barriers around tlb_flush_pending Reading tlb_flush_pending while the page-table lock is taken does not require a barrier, since the lock/unlock already acts as a barrier. Removing the barrier in mm_tlb_flush_pending() to address this issue. However, migrate_misplaced_transhuge_page() calls mm_tlb_flush_pending() while the page-table lock is already released, which may present a problem on architectures with weak memory model (PPC). To deal with this case, a new parameter is added to mm_tlb_flush_pending() to indicate if it is read without the page-table lock taken, and calling smp_mb__after_unlock_lock() in this case. Link: http://lkml.kernel.org/r/20170802000818.4760-3-namit@vmware.com Signed-off-by: Nadav Amit Acked-by: Rik van Riel Cc: Minchan Kim Cc: Sergey Senozhatsky Cc: Andy Lutomirski Cc: Mel Gorman Cc: "David S. Miller" Cc: Andrea Arcangeli Cc: Heiko Carstens Cc: Hugh Dickins Cc: Ingo Molnar Cc: Jeff Dike Cc: Martin Schwidefsky Cc: Mel Gorman Cc: Nadav Amit Cc: Russell King Cc: Tony Luck Cc: Yoshinori Sato Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/mm_types.h | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'include/linux') diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h index f58f76ee1dfa..0e478ebd2706 100644 --- a/include/linux/mm_types.h +++ b/include/linux/mm_types.h @@ -526,12 +526,12 @@ static inline cpumask_t *mm_cpumask(struct mm_struct *mm) /* * Memory barriers to keep this state in sync are graciously provided by * the page table locks, outside of which no page table modifications happen. - * The barriers below prevent the compiler from re-ordering the instructions - * around the memory barriers that are already present in the code. + * The barriers are used to ensure the order between tlb_flush_pending updates, + * which happen while the lock is not taken, and the PTE updates, which happen + * while the lock is taken, are serialized. */ static inline bool mm_tlb_flush_pending(struct mm_struct *mm) { - barrier(); return atomic_read(&mm->tlb_flush_pending) > 0; } @@ -554,7 +554,13 @@ static inline void inc_tlb_flush_pending(struct mm_struct *mm) /* Clearing is done after a TLB flush, which also provides a barrier. */ static inline void dec_tlb_flush_pending(struct mm_struct *mm) { - barrier(); + /* + * Guarantee that the tlb_flush_pending does not not leak into the + * critical section, since we must order the PTE change and changes to + * the pending TLB flush indication. We could have relied on TLB flush + * as a memory barrier, but this behavior is not clearly documented. + */ + smp_mb__before_atomic(); atomic_dec(&mm->tlb_flush_pending); } #else -- cgit v1.2.3 From 56236a59556cfd3bae7bffb7e5f438b5ef0af880 Mon Sep 17 00:00:00 2001 From: Minchan Kim Date: Thu, 10 Aug 2017 15:24:05 -0700 Subject: mm: refactor TLB gathering API This patch is a preparatory patch for solving race problems caused by TLB batch. For that, we will increase/decrease TLB flush pending count of mm_struct whenever tlb_[gather|finish]_mmu is called. Before making it simple, this patch separates architecture specific part and rename it to arch_tlb_[gather|finish]_mmu and generic part just calls it. It shouldn't change any behavior. Link: http://lkml.kernel.org/r/20170802000818.4760-5-namit@vmware.com Signed-off-by: Minchan Kim Signed-off-by: Nadav Amit Acked-by: Mel Gorman Cc: Ingo Molnar Cc: Russell King Cc: Tony Luck Cc: Martin Schwidefsky Cc: "David S. Miller" Cc: Heiko Carstens Cc: Yoshinori Sato Cc: Jeff Dike Cc: Andrea Arcangeli Cc: Andy Lutomirski Cc: Hugh Dickins Cc: Mel Gorman Cc: Nadav Amit Cc: Rik van Riel Cc: Sergey Senozhatsky Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- arch/arm/include/asm/tlb.h | 6 ++++-- arch/ia64/include/asm/tlb.h | 6 ++++-- arch/s390/include/asm/tlb.h | 12 ++++++------ arch/sh/include/asm/tlb.h | 6 ++++-- arch/um/include/asm/tlb.h | 8 +++++--- include/asm-generic/tlb.h | 7 ++++--- include/linux/mm_types.h | 6 ++++++ mm/memory.c | 28 +++++++++++++++++++++------- 8 files changed, 54 insertions(+), 25 deletions(-) (limited to 'include/linux') diff --git a/arch/arm/include/asm/tlb.h b/arch/arm/include/asm/tlb.h index 3f2eb76243e3..7f5b2a2d3861 100644 --- a/arch/arm/include/asm/tlb.h +++ b/arch/arm/include/asm/tlb.h @@ -148,7 +148,8 @@ static inline void tlb_flush_mmu(struct mmu_gather *tlb) } static inline void -tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end) +arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end) { tlb->mm = mm; tlb->fullmm = !(start | (end+1)); @@ -166,7 +167,8 @@ tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start } static inline void -tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end) +arch_tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end) { tlb_flush_mmu(tlb); diff --git a/arch/ia64/include/asm/tlb.h b/arch/ia64/include/asm/tlb.h index fced197b9626..93cadc04ac62 100644 --- a/arch/ia64/include/asm/tlb.h +++ b/arch/ia64/include/asm/tlb.h @@ -168,7 +168,8 @@ static inline void __tlb_alloc_page(struct mmu_gather *tlb) static inline void -tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end) +arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end) { tlb->mm = mm; tlb->max = ARRAY_SIZE(tlb->local); @@ -185,7 +186,8 @@ tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start * collected. */ static inline void -tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end) +arch_tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end) { /* * Note: tlb->nr may be 0 at this point, so we can't rely on tlb->start_addr and diff --git a/arch/s390/include/asm/tlb.h b/arch/s390/include/asm/tlb.h index 7317b3108a88..d574d0820dc8 100644 --- a/arch/s390/include/asm/tlb.h +++ b/arch/s390/include/asm/tlb.h @@ -47,10 +47,9 @@ struct mmu_table_batch { extern void tlb_table_flush(struct mmu_gather *tlb); extern void tlb_remove_table(struct mmu_gather *tlb, void *table); -static inline void tlb_gather_mmu(struct mmu_gather *tlb, - struct mm_struct *mm, - unsigned long start, - unsigned long end) +static inline void +arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end) { tlb->mm = mm; tlb->start = start; @@ -76,8 +75,9 @@ static inline void tlb_flush_mmu(struct mmu_gather *tlb) tlb_flush_mmu_free(tlb); } -static inline void tlb_finish_mmu(struct mmu_gather *tlb, - unsigned long start, unsigned long end) +static inline void +arch_tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end) { tlb_flush_mmu(tlb); } diff --git a/arch/sh/include/asm/tlb.h b/arch/sh/include/asm/tlb.h index 46e0d635e36f..89786560dbd4 100644 --- a/arch/sh/include/asm/tlb.h +++ b/arch/sh/include/asm/tlb.h @@ -36,7 +36,8 @@ static inline void init_tlb_gather(struct mmu_gather *tlb) } static inline void -tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end) +arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end) { tlb->mm = mm; tlb->start = start; @@ -47,7 +48,8 @@ tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start } static inline void -tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end) +arch_tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end) { if (tlb->fullmm) flush_tlb_mm(tlb->mm); diff --git a/arch/um/include/asm/tlb.h b/arch/um/include/asm/tlb.h index 600a2e9bfee2..2a901eca7145 100644 --- a/arch/um/include/asm/tlb.h +++ b/arch/um/include/asm/tlb.h @@ -45,7 +45,8 @@ static inline void init_tlb_gather(struct mmu_gather *tlb) } static inline void -tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end) +arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end) { tlb->mm = mm; tlb->start = start; @@ -80,12 +81,13 @@ tlb_flush_mmu(struct mmu_gather *tlb) tlb_flush_mmu_free(tlb); } -/* tlb_finish_mmu +/* arch_tlb_finish_mmu * Called at the end of the shootdown operation to free up any resources * that were required. */ static inline void -tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end) +arch_tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end) { tlb_flush_mmu(tlb); diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h index 8afa4335e5b2..8f71521e7a44 100644 --- a/include/asm-generic/tlb.h +++ b/include/asm-generic/tlb.h @@ -112,10 +112,11 @@ struct mmu_gather { #define HAVE_GENERIC_MMU_GATHER -void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end); +void arch_tlb_gather_mmu(struct mmu_gather *tlb, + struct mm_struct *mm, unsigned long start, unsigned long end); void tlb_flush_mmu(struct mmu_gather *tlb); -void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, - unsigned long end); +void arch_tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end); extern bool __tlb_remove_page_size(struct mmu_gather *tlb, struct page *page, int page_size); diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h index 0e478ebd2706..c605f2a3a68e 100644 --- a/include/linux/mm_types.h +++ b/include/linux/mm_types.h @@ -522,6 +522,12 @@ static inline cpumask_t *mm_cpumask(struct mm_struct *mm) return mm->cpu_vm_mask_var; } +struct mmu_gather; +extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end); +extern void tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end); + #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) /* * Memory barriers to keep this state in sync are graciously provided by diff --git a/mm/memory.c b/mm/memory.c index f65beaad319b..34cba5113e06 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -215,12 +215,8 @@ static bool tlb_next_batch(struct mmu_gather *tlb) return true; } -/* tlb_gather_mmu - * Called to initialize an (on-stack) mmu_gather structure for page-table - * tear-down from @mm. The @fullmm argument is used when @mm is without - * users and we're going to destroy the full address space (exit/execve). - */ -void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end) +void arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end) { tlb->mm = mm; @@ -275,7 +271,8 @@ void tlb_flush_mmu(struct mmu_gather *tlb) * Called at the end of the shootdown operation to free up any resources * that were required. */ -void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end) +void arch_tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end) { struct mmu_gather_batch *batch, *next; @@ -398,6 +395,23 @@ void tlb_remove_table(struct mmu_gather *tlb, void *table) #endif /* CONFIG_HAVE_RCU_TABLE_FREE */ +/* tlb_gather_mmu + * Called to initialize an (on-stack) mmu_gather structure for page-table + * tear-down from @mm. The @fullmm argument is used when @mm is without + * users and we're going to destroy the full address space (exit/execve). + */ +void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, + unsigned long start, unsigned long end) +{ + arch_tlb_gather_mmu(tlb, mm, start, end); +} + +void tlb_finish_mmu(struct mmu_gather *tlb, + unsigned long start, unsigned long end) +{ + arch_tlb_finish_mmu(tlb, start, end); +} + /* * Note: this doesn't free the actual pages themselves. That * has been handled earlier when unmapping all the memory regions. -- cgit v1.2.3 From 0a2dd266dd6b7a31503b5bbe63af05961a6b446d Mon Sep 17 00:00:00 2001 From: Minchan Kim Date: Thu, 10 Aug 2017 15:24:09 -0700 Subject: mm: make tlb_flush_pending global Currently, tlb_flush_pending is used only for CONFIG_[NUMA_BALANCING| COMPACTION] but upcoming patches to solve subtle TLB flush batching problem will use it regardless of compaction/NUMA so this patch doesn't remove the dependency. [akpm@linux-foundation.org: remove more ifdefs from world's ugliest printk statement] Link: http://lkml.kernel.org/r/20170802000818.4760-6-namit@vmware.com Signed-off-by: Minchan Kim Signed-off-by: Nadav Amit Acked-by: Mel Gorman Cc: "David S. Miller" Cc: Andrea Arcangeli Cc: Andy Lutomirski Cc: Heiko Carstens Cc: Hugh Dickins Cc: Ingo Molnar Cc: Jeff Dike Cc: Martin Schwidefsky Cc: Mel Gorman Cc: Nadav Amit Cc: Rik van Riel Cc: Russell King Cc: Sergey Senozhatsky Cc: Tony Luck Cc: Yoshinori Sato Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/mm_types.h | 21 --------------------- mm/debug.c | 4 ---- 2 files changed, 25 deletions(-) (limited to 'include/linux') diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h index c605f2a3a68e..892a7b0196fd 100644 --- a/include/linux/mm_types.h +++ b/include/linux/mm_types.h @@ -487,14 +487,12 @@ struct mm_struct { /* numa_scan_seq prevents two threads setting pte_numa */ int numa_scan_seq; #endif -#if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) /* * An operation with batched TLB flushing is going on. Anything that * can move process memory needs to flush the TLB when moving a * PROT_NONE or PROT_NUMA mapped page. */ atomic_t tlb_flush_pending; -#endif #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH /* See flush_tlb_batched_pending() */ bool tlb_flush_batched; @@ -528,7 +526,6 @@ extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, extern void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end); -#if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) /* * Memory barriers to keep this state in sync are graciously provided by * the page table locks, outside of which no page table modifications happen. @@ -569,24 +566,6 @@ static inline void dec_tlb_flush_pending(struct mm_struct *mm) smp_mb__before_atomic(); atomic_dec(&mm->tlb_flush_pending); } -#else -static inline bool mm_tlb_flush_pending(struct mm_struct *mm) -{ - return false; -} - -static inline void init_tlb_flush_pending(struct mm_struct *mm) -{ -} - -static inline void inc_tlb_flush_pending(struct mm_struct *mm) -{ -} - -static inline void dec_tlb_flush_pending(struct mm_struct *mm) -{ -} -#endif struct vm_fault; diff --git a/mm/debug.c b/mm/debug.c index d70103bb4731..5715448ab0b5 100644 --- a/mm/debug.c +++ b/mm/debug.c @@ -124,9 +124,7 @@ void dump_mm(const struct mm_struct *mm) #ifdef CONFIG_NUMA_BALANCING "numa_next_scan %lu numa_scan_offset %lu numa_scan_seq %d\n" #endif -#if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) "tlb_flush_pending %d\n" -#endif "def_flags: %#lx(%pGv)\n", mm, mm->mmap, mm->vmacache_seqnum, mm->task_size, @@ -158,9 +156,7 @@ void dump_mm(const struct mm_struct *mm) #ifdef CONFIG_NUMA_BALANCING mm->numa_next_scan, mm->numa_scan_offset, mm->numa_scan_seq, #endif -#if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) atomic_read(&mm->tlb_flush_pending), -#endif mm->def_flags, &mm->def_flags ); } -- cgit v1.2.3 From 99baac21e4585f4258f919502c6e23f1e5edc98c Mon Sep 17 00:00:00 2001 From: Minchan Kim Date: Thu, 10 Aug 2017 15:24:12 -0700 Subject: mm: fix MADV_[FREE|DONTNEED] TLB flush miss problem Nadav reported parallel MADV_DONTNEED on same range has a stale TLB problem and Mel fixed it[1] and found same problem on MADV_FREE[2]. Quote from Mel Gorman: "The race in question is CPU 0 running madv_free and updating some PTEs while CPU 1 is also running madv_free and looking at the same PTEs. CPU 1 may have writable TLB entries for a page but fail the pte_dirty check (because CPU 0 has updated it already) and potentially fail to flush. Hence, when madv_free on CPU 1 returns, there are still potentially writable TLB entries and the underlying PTE is still present so that a subsequent write does not necessarily propagate the dirty bit to the underlying PTE any more. Reclaim at some unknown time at the future may then see that the PTE is still clean and discard the page even though a write has happened in the meantime. I think this is possible but I could have missed some protection in madv_free that prevents it happening." This patch aims for solving both problems all at once and is ready for other problem with KSM, MADV_FREE and soft-dirty story[3]. TLB batch API(tlb_[gather|finish]_mmu] uses [inc|dec]_tlb_flush_pending and mmu_tlb_flush_pending so that when tlb_finish_mmu is called, we can catch there are parallel threads going on. In that case, forcefully, flush TLB to prevent for user to access memory via stale TLB entry although it fail to gather page table entry. I confirmed this patch works with [4] test program Nadav gave so this patch supersedes "mm: Always flush VMA ranges affected by zap_page_range v2" in current mmotm. NOTE: This patch modifies arch-specific TLB gathering interface(x86, ia64, s390, sh, um). It seems most of architecture are straightforward but s390 need to be careful because tlb_flush_mmu works only if mm->context.flush_mm is set to non-zero which happens only a pte entry really is cleared by ptep_get_and_clear and friends. However, this problem never changes the pte entries but need to flush to prevent memory access from stale tlb. [1] http://lkml.kernel.org/r/20170725101230.5v7gvnjmcnkzzql3@techsingularity.net [2] http://lkml.kernel.org/r/20170725100722.2dxnmgypmwnrfawp@suse.de [3] http://lkml.kernel.org/r/BD3A0EBE-ECF4-41D4-87FA-C755EA9AB6BD@gmail.com [4] https://patchwork.kernel.org/patch/9861621/ [minchan@kernel.org: decrease tlb flush pending count in tlb_finish_mmu] Link: http://lkml.kernel.org/r/20170808080821.GA31730@bbox Link: http://lkml.kernel.org/r/20170802000818.4760-7-namit@vmware.com Signed-off-by: Minchan Kim Signed-off-by: Nadav Amit Reported-by: Nadav Amit Reported-by: Mel Gorman Acked-by: Mel Gorman Cc: Ingo Molnar Cc: Russell King Cc: Tony Luck Cc: Martin Schwidefsky Cc: "David S. Miller" Cc: Heiko Carstens Cc: Yoshinori Sato Cc: Jeff Dike Cc: Andrea Arcangeli Cc: Andy Lutomirski Cc: Hugh Dickins Cc: Mel Gorman Cc: Nadav Amit Cc: Rik van Riel Cc: Sergey Senozhatsky Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- arch/arm/include/asm/tlb.h | 7 ++++++- arch/ia64/include/asm/tlb.h | 4 +++- arch/s390/include/asm/tlb.h | 7 ++++++- arch/sh/include/asm/tlb.h | 4 ++-- arch/um/include/asm/tlb.h | 7 ++++++- include/asm-generic/tlb.h | 2 +- include/linux/mm_types.h | 8 ++++++++ mm/memory.c | 18 ++++++++++++++++-- 8 files changed, 48 insertions(+), 9 deletions(-) (limited to 'include/linux') diff --git a/arch/arm/include/asm/tlb.h b/arch/arm/include/asm/tlb.h index 7f5b2a2d3861..d5562f9ce600 100644 --- a/arch/arm/include/asm/tlb.h +++ b/arch/arm/include/asm/tlb.h @@ -168,8 +168,13 @@ arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, static inline void arch_tlb_finish_mmu(struct mmu_gather *tlb, - unsigned long start, unsigned long end) + unsigned long start, unsigned long end, bool force) { + if (force) { + tlb->range_start = start; + tlb->range_end = end; + } + tlb_flush_mmu(tlb); /* keep the page table cache within bounds */ diff --git a/arch/ia64/include/asm/tlb.h b/arch/ia64/include/asm/tlb.h index 93cadc04ac62..cbe5ac3699bf 100644 --- a/arch/ia64/include/asm/tlb.h +++ b/arch/ia64/include/asm/tlb.h @@ -187,8 +187,10 @@ arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, */ static inline void arch_tlb_finish_mmu(struct mmu_gather *tlb, - unsigned long start, unsigned long end) + unsigned long start, unsigned long end, bool force) { + if (force) + tlb->need_flush = 1; /* * Note: tlb->nr may be 0 at this point, so we can't rely on tlb->start_addr and * tlb->end_addr. diff --git a/arch/s390/include/asm/tlb.h b/arch/s390/include/asm/tlb.h index d574d0820dc8..2eb8ff0d6fca 100644 --- a/arch/s390/include/asm/tlb.h +++ b/arch/s390/include/asm/tlb.h @@ -77,8 +77,13 @@ static inline void tlb_flush_mmu(struct mmu_gather *tlb) static inline void arch_tlb_finish_mmu(struct mmu_gather *tlb, - unsigned long start, unsigned long end) + unsigned long start, unsigned long end, bool force) { + if (force) { + tlb->start = start; + tlb->end = end; + } + tlb_flush_mmu(tlb); } diff --git a/arch/sh/include/asm/tlb.h b/arch/sh/include/asm/tlb.h index 89786560dbd4..51a8bc967e75 100644 --- a/arch/sh/include/asm/tlb.h +++ b/arch/sh/include/asm/tlb.h @@ -49,9 +49,9 @@ arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, static inline void arch_tlb_finish_mmu(struct mmu_gather *tlb, - unsigned long start, unsigned long end) + unsigned long start, unsigned long end, bool force) { - if (tlb->fullmm) + if (tlb->fullmm || force) flush_tlb_mm(tlb->mm); /* keep the page table cache within bounds */ diff --git a/arch/um/include/asm/tlb.h b/arch/um/include/asm/tlb.h index 2a901eca7145..344d95619d03 100644 --- a/arch/um/include/asm/tlb.h +++ b/arch/um/include/asm/tlb.h @@ -87,8 +87,13 @@ tlb_flush_mmu(struct mmu_gather *tlb) */ static inline void arch_tlb_finish_mmu(struct mmu_gather *tlb, - unsigned long start, unsigned long end) + unsigned long start, unsigned long end, bool force) { + if (force) { + tlb->start = start; + tlb->end = end; + tlb->need_flush = 1; + } tlb_flush_mmu(tlb); /* keep the page table cache within bounds */ diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h index 8f71521e7a44..faddde44de8c 100644 --- a/include/asm-generic/tlb.h +++ b/include/asm-generic/tlb.h @@ -116,7 +116,7 @@ void arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end); void tlb_flush_mmu(struct mmu_gather *tlb); void arch_tlb_finish_mmu(struct mmu_gather *tlb, - unsigned long start, unsigned long end); + unsigned long start, unsigned long end, bool force); extern bool __tlb_remove_page_size(struct mmu_gather *tlb, struct page *page, int page_size); diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h index 892a7b0196fd..3cadee0a3508 100644 --- a/include/linux/mm_types.h +++ b/include/linux/mm_types.h @@ -538,6 +538,14 @@ static inline bool mm_tlb_flush_pending(struct mm_struct *mm) return atomic_read(&mm->tlb_flush_pending) > 0; } +/* + * Returns true if there are two above TLB batching threads in parallel. + */ +static inline bool mm_tlb_flush_nested(struct mm_struct *mm) +{ + return atomic_read(&mm->tlb_flush_pending) > 1; +} + static inline void init_tlb_flush_pending(struct mm_struct *mm) { atomic_set(&mm->tlb_flush_pending, 0); diff --git a/mm/memory.c b/mm/memory.c index 34cba5113e06..e158f7ac6730 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -272,10 +272,13 @@ void tlb_flush_mmu(struct mmu_gather *tlb) * that were required. */ void arch_tlb_finish_mmu(struct mmu_gather *tlb, - unsigned long start, unsigned long end) + unsigned long start, unsigned long end, bool force) { struct mmu_gather_batch *batch, *next; + if (force) + __tlb_adjust_range(tlb, start, end - start); + tlb_flush_mmu(tlb); /* keep the page table cache within bounds */ @@ -404,12 +407,23 @@ void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end) { arch_tlb_gather_mmu(tlb, mm, start, end); + inc_tlb_flush_pending(tlb->mm); } void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end) { - arch_tlb_finish_mmu(tlb, start, end); + /* + * If there are parallel threads are doing PTE changes on same range + * under non-exclusive lock(e.g., mmap_sem read-side) but defer TLB + * flush by batching, a thread has stable TLB entry can fail to flush + * the TLB by observing pte_none|!pte_dirty, for example so flush TLB + * forcefully if we detect parallel PTE batching threads. + */ + bool force = mm_tlb_flush_nested(tlb->mm); + + arch_tlb_finish_mmu(tlb, start, end, force); + dec_tlb_flush_pending(tlb->mm); } /* -- cgit v1.2.3