diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2013-02-28 13:16:07 -0800 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2013-02-28 13:16:07 -0800 |
commit | f042fea0da78d9dc077a9c736b33b60eb8f35195 (patch) | |
tree | fce1c64a68a8ccd54b36e4d89d3b946a995e9fd2 | |
parent | ee89f81252179dcbf6cd65bd48299f5e52292d88 (diff) | |
parent | b7a1da695f3fb33d093e6de20b1dfc238e3c9587 (diff) |
Merge branch 'for-3.9/drivers' of git://git.kernel.dk/linux-block
Pull block driver bits from Jens Axboe:
"After the block IO core bits are in, please grab the driver updates
from below as well. It contains:
- Fix ancient regression in dac960. Nobody must be using that
anymore...
- Some good fixes from Guo Ghao for loop, fixing both potential
oopses and deadlocks.
- Improve mtip32xx for NUMA systems, by being a bit more clever in
distributing work.
- Add IBM RamSan 70/80 driver. A second round of fixes for that is
pending, that will come in through for-linus during the 3.9 cycle
as per usual.
- A few xen-blk{back,front} fixes from Konrad and Roger.
- Other minor fixes and improvements."
* 'for-3.9/drivers' of git://git.kernel.dk/linux-block:
loopdev: ignore negative offset when calculate loop device size
loopdev: remove an user triggerable oops
loopdev: move common code into loop_figure_size()
loopdev: update block device size in loop_set_status()
loopdev: fix a deadlock
xen-blkback: use balloon pages for persistent grants
xen-blkfront: drop the use of llist_for_each_entry_safe
xen/blkback: Don't trust the handle from the frontend.
xen-blkback: do not leak mode property
block: IBM RamSan 70/80 driver fixes
rsxx: add slab.h include to dma.c
drivers/block/mtip32xx: add missing GENERIC_HARDIRQS dependency
block: remove new __devinit/exit annotations on ramsam driver
block: IBM RamSan 70/80 device driver
drivers/block/mtip32xx/mtip32xx.c:1726:5: sparse: symbol 'mtip_send_trim' was not declared. Should it be static?
drivers/block/mtip32xx/mtip32xx.c:4029:1: sparse: symbol 'mtip_workq_sdbf0' was not declared. Should it be static?
dac960: return success instead of -ENOTTY
mtip32xx: add trim support
mtip32xx: Add workqueue and NUMA support
block: delete super ancient PC-XT driver for 1980's hardware
-rw-r--r-- | MAINTAINERS | 6 | ||||
-rw-r--r-- | drivers/block/DAC960.c | 1 | ||||
-rw-r--r-- | drivers/block/Kconfig | 23 | ||||
-rw-r--r-- | drivers/block/Makefile | 3 | ||||
-rw-r--r-- | drivers/block/loop.c | 61 | ||||
-rw-r--r-- | drivers/block/mtip32xx/Kconfig | 2 | ||||
-rw-r--r-- | drivers/block/mtip32xx/mtip32xx.c | 431 | ||||
-rw-r--r-- | drivers/block/mtip32xx/mtip32xx.h | 48 | ||||
-rw-r--r-- | drivers/block/rsxx/Makefile | 2 | ||||
-rw-r--r-- | drivers/block/rsxx/config.c | 213 | ||||
-rw-r--r-- | drivers/block/rsxx/core.c | 649 | ||||
-rw-r--r-- | drivers/block/rsxx/cregs.c | 758 | ||||
-rw-r--r-- | drivers/block/rsxx/dev.c | 367 | ||||
-rw-r--r-- | drivers/block/rsxx/dma.c | 998 | ||||
-rw-r--r-- | drivers/block/rsxx/rsxx.h | 45 | ||||
-rw-r--r-- | drivers/block/rsxx/rsxx_cfg.h | 72 | ||||
-rw-r--r-- | drivers/block/rsxx/rsxx_priv.h | 399 | ||||
-rw-r--r-- | drivers/block/xd.c | 1123 | ||||
-rw-r--r-- | drivers/block/xd.h | 134 | ||||
-rw-r--r-- | drivers/block/xen-blkback/blkback.c | 7 | ||||
-rw-r--r-- | drivers/block/xen-blkback/xenbus.c | 49 | ||||
-rw-r--r-- | drivers/block/xen-blkfront.c | 13 | ||||
-rw-r--r-- | include/linux/llist.h | 25 |
23 files changed, 3986 insertions, 1443 deletions
diff --git a/MAINTAINERS b/MAINTAINERS index 0ade02d5ee6..c7b9b2bf8a0 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -6515,6 +6515,12 @@ S: Maintained F: Documentation/blockdev/ramdisk.txt F: drivers/block/brd.c +RAMSAM DRIVER (IBM RamSan 70/80 PCI SSD Flash Card) +M: Joshua Morris <josh.h.morris@us.ibm.com> +M: Philip Kelleher <pjk1939@linux.vnet.ibm.com> +S: Maintained +F: drivers/block/rsxx/ + RANDOM NUMBER DRIVER M: Theodore Ts'o" <tytso@mit.edu> S: Maintained diff --git a/drivers/block/DAC960.c b/drivers/block/DAC960.c index 8f12dc78a84..5b5ee79ff23 100644 --- a/drivers/block/DAC960.c +++ b/drivers/block/DAC960.c @@ -7054,6 +7054,7 @@ static long DAC960_gam_ioctl(struct file *file, unsigned int Request, else ErrorCode = 0; } + break; default: ErrorCode = -ENOTTY; } diff --git a/drivers/block/Kconfig b/drivers/block/Kconfig index 824e09c4d0d..5dc0daed8fa 100644 --- a/drivers/block/Kconfig +++ b/drivers/block/Kconfig @@ -63,19 +63,6 @@ config AMIGA_Z2RAM To compile this driver as a module, choose M here: the module will be called z2ram. -config BLK_DEV_XD - tristate "XT hard disk support" - depends on ISA && ISA_DMA_API - select CHECK_SIGNATURE - help - Very old 8 bit hard disk controllers used in the IBM XT computer - will be supported if you say Y here. - - To compile this driver as a module, choose M here: the - module will be called xd. - - It's pretty unlikely that you have one of these: say N. - config GDROM tristate "SEGA Dreamcast GD-ROM drive" depends on SH_DREAMCAST @@ -544,4 +531,14 @@ config BLK_DEV_RBD If unsure, say N. +config BLK_DEV_RSXX + tristate "RamSam PCIe Flash SSD Device Driver" + depends on PCI + help + Device driver for IBM's high speed PCIe SSD + storage devices: RamSan-70 and RamSan-80. + + To compile this driver as a module, choose M here: the + module will be called rsxx. + endif # BLK_DEV diff --git a/drivers/block/Makefile b/drivers/block/Makefile index 17e82df3df7..a3b40232c6a 100644 --- a/drivers/block/Makefile +++ b/drivers/block/Makefile @@ -15,7 +15,6 @@ obj-$(CONFIG_ATARI_FLOPPY) += ataflop.o obj-$(CONFIG_AMIGA_Z2RAM) += z2ram.o obj-$(CONFIG_BLK_DEV_RAM) += brd.o obj-$(CONFIG_BLK_DEV_LOOP) += loop.o -obj-$(CONFIG_BLK_DEV_XD) += xd.o obj-$(CONFIG_BLK_CPQ_DA) += cpqarray.o obj-$(CONFIG_BLK_CPQ_CISS_DA) += cciss.o obj-$(CONFIG_BLK_DEV_DAC960) += DAC960.o @@ -41,4 +40,6 @@ obj-$(CONFIG_BLK_DEV_DRBD) += drbd/ obj-$(CONFIG_BLK_DEV_RBD) += rbd.o obj-$(CONFIG_BLK_DEV_PCIESSD_MTIP32XX) += mtip32xx/ +obj-$(CONFIG_BLK_DEV_RSXX) += rsxx/ + swim_mod-y := swim.o swim_asm.o diff --git a/drivers/block/loop.c b/drivers/block/loop.c index f47dccbda1d..747bb2af69d 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -162,12 +162,13 @@ static struct loop_func_table *xfer_funcs[MAX_LO_CRYPT] = { static loff_t get_size(loff_t offset, loff_t sizelimit, struct file *file) { - loff_t size, loopsize; + loff_t loopsize; /* Compute loopsize in bytes */ - size = i_size_read(file->f_mapping->host); - loopsize = size - offset; - /* offset is beyond i_size, wierd but possible */ + loopsize = i_size_read(file->f_mapping->host); + if (offset > 0) + loopsize -= offset; + /* offset is beyond i_size, weird but possible */ if (loopsize < 0) return 0; @@ -190,6 +191,7 @@ figure_loop_size(struct loop_device *lo, loff_t offset, loff_t sizelimit) { loff_t size = get_size(offset, sizelimit, lo->lo_backing_file); sector_t x = (sector_t)size; + struct block_device *bdev = lo->lo_device; if (unlikely((loff_t)x != size)) return -EFBIG; @@ -198,6 +200,9 @@ figure_loop_size(struct loop_device *lo, loff_t offset, loff_t sizelimit) if (lo->lo_sizelimit != sizelimit) lo->lo_sizelimit = sizelimit; set_capacity(lo->lo_disk, x); + bd_set_size(bdev, (loff_t)get_capacity(bdev->bd_disk) << 9); + /* let user-space know about the new size */ + kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); return 0; } @@ -1091,10 +1096,10 @@ loop_set_status(struct loop_device *lo, const struct loop_info64 *info) return err; if (lo->lo_offset != info->lo_offset || - lo->lo_sizelimit != info->lo_sizelimit) { + lo->lo_sizelimit != info->lo_sizelimit) if (figure_loop_size(lo, info->lo_offset, info->lo_sizelimit)) return -EFBIG; - } + loop_config_discard(lo); memcpy(lo->lo_file_name, info->lo_file_name, LO_NAME_SIZE); @@ -1271,28 +1276,10 @@ loop_get_status64(struct loop_device *lo, struct loop_info64 __user *arg) { static int loop_set_capacity(struct loop_device *lo, struct block_device *bdev) { - int err; - sector_t sec; - loff_t sz; - - err = -ENXIO; if (unlikely(lo->lo_state != Lo_bound)) - goto out; - err = figure_loop_size(lo, lo->lo_offset, lo->lo_sizelimit); - if (unlikely(err)) - goto out; - sec = get_capacity(lo->lo_disk); - /* the width of sector_t may be narrow for bit-shift */ - sz = sec; - sz <<= 9; - mutex_lock(&bdev->bd_mutex); - bd_set_size(bdev, sz); - /* let user-space know about the new size */ - kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE); - mutex_unlock(&bdev->bd_mutex); + return -ENXIO; - out: - return err; + return figure_loop_size(lo, lo->lo_offset, lo->lo_sizelimit); } static int lo_ioctl(struct block_device *bdev, fmode_t mode, @@ -1845,11 +1832,15 @@ static int __init loop_init(void) max_part = (1UL << part_shift) - 1; } - if ((1UL << part_shift) > DISK_MAX_PARTS) - return -EINVAL; + if ((1UL << part_shift) > DISK_MAX_PARTS) { + err = -EINVAL; + goto misc_out; + } - if (max_loop > 1UL << (MINORBITS - part_shift)) - return -EINVAL; + if (max_loop > 1UL << (MINORBITS - part_shift)) { + err = -EINVAL; + goto misc_out; + } /* * If max_loop is specified, create that many devices upfront. @@ -1867,8 +1858,10 @@ static int __init loop_init(void) range = 1UL << MINORBITS; } - if (register_blkdev(LOOP_MAJOR, "loop")) - return -EIO; + if (register_blkdev(LOOP_MAJOR, "loop")) { + err = -EIO; + goto misc_out; + } blk_register_region(MKDEV(LOOP_MAJOR, 0), range, THIS_MODULE, loop_probe, NULL, NULL); @@ -1881,6 +1874,10 @@ static int __init loop_init(void) printk(KERN_INFO "loop: module loaded\n"); return 0; + +misc_out: + misc_deregister(&loop_misc); + return err; } static int loop_exit_cb(int id, void *ptr, void *data) diff --git a/drivers/block/mtip32xx/Kconfig b/drivers/block/mtip32xx/Kconfig index 0ba837fc62a..1fca1f996b4 100644 --- a/drivers/block/mtip32xx/Kconfig +++ b/drivers/block/mtip32xx/Kconfig @@ -4,6 +4,6 @@ config BLK_DEV_PCIESSD_MTIP32XX tristate "Block Device Driver for Micron PCIe SSDs" - depends on PCI + depends on PCI && GENERIC_HARDIRQS help This enables the block driver for Micron PCIe SSDs. diff --git a/drivers/block/mtip32xx/mtip32xx.c b/drivers/block/mtip32xx/mtip32xx.c index 3fd10099045..11cc9522cdd 100644 --- a/drivers/block/mtip32xx/mtip32xx.c +++ b/drivers/block/mtip32xx/mtip32xx.c @@ -88,6 +88,8 @@ static int instance; static int mtip_major; static struct dentry *dfs_parent; +static u32 cpu_use[NR_CPUS]; + static DEFINE_SPINLOCK(rssd_index_lock); static DEFINE_IDA(rssd_index_ida); @@ -296,16 +298,17 @@ static int hba_reset_nosleep(struct driver_data *dd) */ static inline void mtip_issue_ncq_command(struct mtip_port *port, int tag) { - atomic_set(&port->commands[tag].active, 1); + int group = tag >> 5; - spin_lock(&port->cmd_issue_lock); + atomic_set(&port->commands[tag].active, 1); + /* guard SACT and CI registers */ + spin_lock(&port->cmd_issue_lock[group]); writel((1 << MTIP_TAG_BIT(tag)), port->s_active[MTIP_TAG_INDEX(tag)]); writel((1 << MTIP_TAG_BIT(tag)), port->cmd_issue[MTIP_TAG_INDEX(tag)]); - - spin_unlock(&port->cmd_issue_lock); + spin_unlock(&port->cmd_issue_lock[group]); /* Set the command's timeout value.*/ port->commands[tag].comp_time = jiffies + msecs_to_jiffies( @@ -964,56 +967,56 @@ handle_tfe_exit: /* * Handle a set device bits interrupt */ -static inline void mtip_process_sdbf(struct driver_data *dd) +static inline void mtip_workq_sdbfx(struct mtip_port *port, int group, + u32 completed) { - struct mtip_port *port = dd->port; - int group, tag, bit; - u32 completed; + struct driver_data *dd = port->dd; + int tag, bit; struct mtip_cmd *command; - /* walk all bits in all slot groups */ - for (group = 0; group < dd->slot_groups; group++) { - completed = readl(port->completed[group]); - if (!completed) - continue; + if (!completed) { + WARN_ON_ONCE(!completed); + return; + } + /* clear completed status register in the hardware.*/ + writel(completed, port->completed[group]); - /* clear completed status register in the hardware.*/ - writel(completed, port->completed[group]); + /* Process completed commands. */ + for (bit = 0; (bit < 32) && completed; bit++) { + if (completed & 0x01) { + tag = (group << 5) | bit; - /* Process completed commands. */ - for (bit = 0; - (bit < 32) && completed; - bit++, completed >>= 1) { - if (completed & 0x01) { - tag = (group << 5) | bit; + /* skip internal command slot. */ + if (unlikely(tag == MTIP_TAG_INTERNAL)) + continue; - /* skip internal command slot. */ - if (unlikely(tag == MTIP_TAG_INTERNAL)) - continue; + command = &port->commands[tag]; + /* make internal callback */ + if (likely(command->comp_func)) { + command->comp_func( + port, + tag, + command->comp_data, + 0); + } else { + dev_warn(&dd->pdev->dev, + "Null completion " + "for tag %d", + tag); - command = &port->commands[tag]; - /* make internal callback */ - if (likely(command->comp_func)) { - command->comp_func( - port, - tag, - command->comp_data, - 0); - } else { - dev_warn(&dd->pdev->dev, - "Null completion " - "for tag %d", - tag); - - if (mtip_check_surprise_removal( - dd->pdev)) { - mtip_command_cleanup(dd); - return; - } + if (mtip_check_surprise_removal( + dd->pdev)) { + mtip_command_cleanup(dd); + return; } } } + completed >>= 1; } + + /* If last, re-enable interrupts */ + if (atomic_dec_return(&dd->irq_workers_active) == 0) + writel(0xffffffff, dd->mmio + HOST_IRQ_STAT); } /* @@ -1072,6 +1075,8 @@ static inline irqreturn_t mtip_handle_irq(struct driver_data *data) struct mtip_port *port = dd->port; u32 hba_stat, port_stat; int rv = IRQ_NONE; + int do_irq_enable = 1, i, workers; + struct mtip_work *twork; hba_stat = readl(dd->mmio + HOST_IRQ_STAT); if (hba_stat) { @@ -1082,8 +1087,42 @@ static inline irqreturn_t mtip_handle_irq(struct driver_data *data) writel(port_stat, port->mmio + PORT_IRQ_STAT); /* Demux port status */ - if (likely(port_stat & PORT_IRQ_SDB_FIS)) - mtip_process_sdbf(dd); + if (likely(port_stat & PORT_IRQ_SDB_FIS)) { + do_irq_enable = 0; + WARN_ON_ONCE(atomic_read(&dd->irq_workers_active) != 0); + + /* Start at 1: group zero is always local? */ + for (i = 0, workers = 0; i < MTIP_MAX_SLOT_GROUPS; + i++) { + twork = &dd->work[i]; + twork->completed = readl(port->completed[i]); + if (twork->completed) + workers++; + } + + atomic_set(&dd->irq_workers_active, workers); + if (workers) { + for (i = 1; i < MTIP_MAX_SLOT_GROUPS; i++) { + twork = &dd->work[i]; + if (twork->completed) + queue_work_on( + twork->cpu_binding, + dd->isr_workq, + &twork->work); + } + + if (likely(dd->work[0].completed)) + mtip_workq_sdbfx(port, 0, + dd->work[0].completed); + + } else { + /* + * Chip quirk: SDB interrupt but nothing + * to complete + */ + do_irq_enable = 1; + } + } if (unlikely(port_stat & PORT_IRQ_ERR)) { if (unlikely(mtip_check_surprise_removal(dd->pdev))) { @@ -1103,21 +1142,13 @@ static inline irqreturn_t mtip_handle_irq(struct driver_data *data) } /* acknowledge interrupt */ - writel(hba_stat, dd->mmio + HOST_IRQ_STAT); + if (unlikely(do_irq_enable)) + writel(hba_stat, dd->mmio + HOST_IRQ_STAT); return rv; } /* - * Wrapper for mtip_handle_irq - * (ignores return code) - */ -static void mtip_tasklet(unsigned long data) -{ - mtip_handle_irq((struct driver_data *) data); -} - -/* * HBA interrupt subroutine. * * @irq IRQ number. @@ -1130,8 +1161,8 @@ static void mtip_tasklet(unsigned long data) static irqreturn_t mtip_irq_handler(int irq, void *instance) { struct driver_data *dd = instance; - tasklet_schedule(&dd->tasklet); - return IRQ_HANDLED; + + return mtip_handle_irq(dd); } static void mtip_issue_non_ncq_command(struct mtip_port *port, int tag) @@ -1489,6 +1520,12 @@ static int mtip_get_identify(struct mtip_port *port, void __user *user_buffer) } #endif + /* Demux ID.DRAT & ID.RZAT to determine trim support */ + if (port->identify[69] & (1 << 14) && port->identify[69] & (1 << 5)) + port->dd->trim_supp = true; + else + port->dd->trim_supp = false; + /* Set the identify buffer as valid. */ port->identify_valid = 1; @@ -1676,6 +1713,81 @@ static int mtip_get_smart_attr(struct mtip_port *port, unsigned int id, } /* + * Trim unused sectors + * + * @dd pointer to driver_data structure + * @lba starting lba + * @len # of 512b sectors to trim + * + * return value + * -ENOMEM Out of dma memory + * -EINVAL Invalid parameters passed in, trim not supported + * -EIO Error submitting trim request to hw + */ +static int mtip_send_trim(struct driver_data *dd, unsigned int lba, unsigned int len) +{ + int i, rv = 0; + u64 tlba, tlen, sect_left; + struct mtip_trim_entry *buf; + dma_addr_t dma_addr; + struct host_to_dev_fis fis; + + if (!len || dd->trim_supp == false) + return -EINVAL; + + /* Trim request too big */ + WARN_ON(len > (MTIP_MAX_TRIM_ENTRY_LEN * MTIP_MAX_TRIM_ENTRIES)); + + /* Trim request not aligned on 4k boundary */ + WARN_ON(len % 8 != 0); + + /* Warn if vu_trim structure is too big */ + WARN_ON(sizeof(struct mtip_trim) > ATA_SECT_SIZE); + + /* Allocate a DMA buffer for the trim structure */ + buf = dmam_alloc_coherent(&dd->pdev->dev, ATA_SECT_SIZE, &dma_addr, + GFP_KERNEL); + if (!buf) + return -ENOMEM; + memset(buf, 0, ATA_SECT_SIZE); + + for (i = 0, sect_left = len, tlba = lba; + i < MTIP_MAX_TRIM_ENTRIES && sect_left; + i++) { + tlen = (sect_left >= MTIP_MAX_TRIM_ENTRY_LEN ? + MTIP_MAX_TRIM_ENTRY_LEN : + sect_left); + buf[i].lba = __force_bit2int cpu_to_le32(tlba); + buf[i].range = __force_bit2int cpu_to_le16(tlen); + tlba += tlen; + sect_left -= tlen; + } + WARN_ON(sect_left != 0); + + /* Build the fis */ + memset(&fis, 0, sizeof(struct host_to_dev_fis)); + fis.type = 0x27; + fis.opts = 1 << 7; + fis.command = 0xfb; + fis.features = 0x60; + fis.sect_count = 1; + fis.device = ATA_DEVICE_OBS; + + if (mtip_exec_internal_command(dd->port, + &fis, + 5, + dma_addr, + ATA_SECT_SIZE, + 0, + GFP_KERNEL, + MTIP_TRIM_TIMEOUT_MS) < 0) + rv = -EIO; + + dmam_free_coherent(&dd->pdev->dev, ATA_SECT_SIZE, buf, dma_addr); + return rv; +} + +/* * Get the drive capacity. * * @dd Pointer to the device data structure. @@ -3005,20 +3117,24 @@ static int mtip_hw_init(struct driver_data *dd) hba_setup(dd); - tasklet_init(&dd->tasklet, mtip_tasklet, (unsigned long)dd); - - dd->port = kzalloc(sizeof(struct mtip_port), GFP_KERNEL); + dd->port = kzalloc_node(sizeof(struct mtip_port), GFP_KERNEL, + dd->numa_node); if (!dd->port) { dev_err(&dd->pdev->dev, "Memory allocation: port structure\n"); return -ENOMEM; } + /* Continue workqueue setup */ + for (i = 0; i < MTIP_MAX_SLOT_GROUPS; i++) + dd->work[i].port = dd->port; + /* Counting semaphore to track command slot usage */ sema_init(&dd->port->cmd_slot, num_command_slots - 1); /* Spinlock to prevent concurrent issue */ - spin_lock_init(&dd->port->cmd_issue_lock); + for (i = 0; i < MTIP_MAX_SLOT_GROUPS; i++) + spin_lock_init(&dd->port->cmd_issue_lock[i]); /* Set the port mmio base address. */ dd->port->mmio = dd->mmio + PORT_OFFSET; @@ -3165,6 +3281,7 @@ static int mtip_hw_init(struct driver_data *dd) "Unable to allocate IRQ %d\n", dd->pdev->irq); goto out2; } + irq_set_affinity_hint(dd->pdev->irq, get_cpu_mask(dd->isr_binding)); /* Enable interrupts on the HBA. */ writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN, @@ -3241,7 +3358,8 @@ out3: writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN, dd->mmio + HOST_CTL); - /*Release the IRQ. */ + /* Release the IRQ. */ + irq_set_affinity_hint(dd->pdev->irq, NULL); devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd); out2: @@ -3291,11 +3409,9 @@ static int mtip_hw_exit(struct driver_data *dd) del_timer_sync(&dd->port->cmd_timer); /* Release the IRQ. */ + irq_set_affinity_hint(dd->pdev->irq, NULL); devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd); - /* Stop the bottom half tasklet. */ - tasklet_kill(&dd->tasklet); - /* Free the command/command header memory. */ dmam_free_coherent(&dd->pdev->dev, HW_PORT_PRIV_DMA_SZ + (ATA_SECT_SIZE * 4), @@ -3641,6 +3757,12 @@ static void mtip_make_request(struct request_queue *queue, struct bio *bio) } } + if (unlikely(bio->bi_rw & REQ_DISCARD)) { + bio_endio(bio, mtip_send_trim(dd, bio->bi_sector, + bio_sectors(bio))); + return; + } + if (unlikely(!bio_has_data(bio))) { blk_queue_flush(queue, 0); bio_endio(bio, 0); @@ -3711,7 +3833,7 @@ static int mtip_block_initialize(struct driver_data *dd) goto protocol_init_error; } - dd->disk = alloc_disk(MTIP_MAX_MINORS); + dd->disk = alloc_disk_node(MTIP_MAX_MINORS, dd->numa_node); if (dd->disk == NULL) { dev_err(&dd->pdev->dev, "Unable to allocate gendisk structure\n"); @@ -3755,7 +3877,7 @@ static int mtip_block_initialize(struct driver_data *dd) skip_create_disk: /* Allocate the request queue. */ - dd->queue = blk_alloc_queue(GFP_KERNEL); + dd->queue = blk_alloc_queue_node(GFP_KERNEL, dd->numa_node); if (dd->queue == NULL) { dev_err(&dd->pdev->dev, "Unable to allocate request queue\n"); @@ -3783,6 +3905,15 @@ skip_create_disk: */ blk_queue_flush(dd->queue, 0); + /* Signal trim support */ + if (dd->trim_supp == true) { + set_bit(QUEUE_FLAG_DISCARD, &dd->queue->queue_flags); + dd->queue->limits.discard_granularity = 4096; + blk_queue_max_discard_sectors(dd->queue, + MTIP_MAX_TRIM_ENTRY_LEN * MTIP_MAX_TRIM_ENTRIES); + dd->queue->limits.discard_zeroes_data = 0; + } + /* Set the capacity of the device in 512 byte sectors. */ if (!(mtip_hw_get_capacity(dd, &capacity))) { dev_warn(&dd->pdev->dev, @@ -3813,9 +3944,8 @@ skip_create_disk: start_service_thread: sprintf(thd_name, "mtip_svc_thd_%02d", index); - - dd->mtip_svc_handler = kthread_run(mtip_service_thread, - dd, thd_name); + dd->mtip_svc_handler = kthread_create_on_node(mtip_service_thread, + dd, dd->numa_node, thd_name); if (IS_ERR(dd->mtip_svc_handler)) { dev_err(&dd->pdev->dev, "service thread failed to start\n"); @@ -3823,7 +3953,7 @@ start_service_thread: rv = -EFAULT; goto kthread_run_error; } - + wake_up_process(dd->mtip_svc_handler); if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC) rv = wait_for_rebuild; @@ -3963,6 +4093,56 @@ static int mtip_block_resume(struct driver_data *dd) return 0; } +static void drop_cpu(int cpu) +{ + cpu_use[cpu]--; +} + +static int get_least_used_cpu_on_node(int node) +{ + int cpu, least_used_cpu, least_cnt; + const struct cpumask *node_mask; + + node_mask = cpumask_of_node(node); + least_used_cpu = cpumask_first(node_mask); + least_cnt = cpu_use[least_used_cpu]; + cpu = least_used_cpu; + + for_each_cpu(cpu, node_mask) { + if (cpu_use[cpu] < least_cnt) { + least_used_cpu = cpu; + least_cnt = cpu_use[cpu]; + } + } + cpu_use[least_used_cpu]++; + return least_used_cpu; +} + +/* Helper for selecting a node in round robin mode */ +static inline int mtip_get_next_rr_node(void) +{ + static int next_node = -1; + + if (next_node == -1) { + next_node = first_online_node; + return next_node; + } + + next_node = next_online_node(next_node); + if (next_node == MAX_NUMNODES) + next_node = first_online_node; + return next_node; +} + +static DEFINE_HANDLER(0); +static DEFINE_HANDLER(1); +static DEFINE_HANDLER(2); +static DEFINE_HANDLER(3); +static DEFINE_HANDLER(4); +static DEFINE_HANDLER(5); +static DEFINE_HANDLER(6); +static DEFINE_HANDLER(7); + /* * Called for each supported PCI device detected. * @@ -3977,9 +4157,25 @@ static int mtip_pci_probe(struct pci_dev *pdev, { int rv = 0; struct driver_data *dd = NULL; + char cpu_list[256]; + const struct cpumask *node_mask; + int cpu, i = 0, j = 0; + int my_node = NUMA_NO_NODE; /* Allocate memory for this devices private data. */ - dd = kzalloc(sizeof(struct driver_data), GFP_KERNEL); + my_node = pcibus_to_node(pdev->bus); + if (my_node != NUMA_NO_NODE) { + if (!node_online(my_node)) + my_node = mtip_get_next_rr_node(); + } else { + dev_info(&pdev->dev, "Kernel not reporting proximity, choosing a node\n"); + my_node = mtip_get_next_rr_node(); + } + dev_info(&pdev->dev, "NUMA node %d (closest: %d,%d, probe on %d:%d)\n", + my_node, pcibus_to_node(pdev->bus), dev_to_node(&pdev->dev), + cpu_to_node(smp_processor_id()), smp_processor_id()); + + dd = kzalloc_node(sizeof(struct driver_data), GFP_KERNEL, my_node); if (dd == NULL) { dev_err(&pdev->dev, "Unable to allocate memory for driver data\n"); @@ -4016,19 +4212,82 @@ static int mtip_pci_probe(struct pci_dev *pdev, } } - pci_set_master(pdev); + /* Copy the info we may need later into the private data structure. */ + dd->major = mtip_major; + dd->instance = instance; + dd->pdev = pdev; + dd->numa_node = my_node; + memset(dd->workq_name, 0, 32); + snprintf(dd->workq_name, 31, "mtipq%d", dd->instance); + + dd->isr_workq = create_workqueue(dd->workq_name); + if (!dd->isr_workq) { + dev_warn(&pdev->dev, "Can't create wq %d\n", dd->instance); + goto block_initialize_err; + } + + memset(cpu_list, 0, sizeof(cpu_list)); + + node_mask = cpumask_of_node(dd->numa_node); + if (!cpumask_empty(node_mask)) { + for_each_cpu(cpu, node_mask) + { + snprintf(&cpu_list[j], 256 - j, "%d ", cpu); + j = strlen(cpu_list); + } + + dev_info(&pdev->dev, "Node %d on package %d has %d cpu(s): %s\n", + dd->numa_node, + topology_physical_package_id(cpumask_first(node_mask)), + nr_cpus_node(dd->numa_node), + cpu_list); + } else + dev_dbg(&pdev->dev, "mtip32xx: node_mask empty\n"); + + dd->isr_binding = get_least_used_cpu_on_node(dd->numa_node); + dev_info(&pdev->dev, "Initial IRQ binding node:cpu %d:%d\n", + cpu_to_node(dd->isr_binding), dd->isr_binding); + + /* first worker context always runs in ISR */ + dd->work[0].cpu_binding = dd->isr_binding; + dd->work[1].cpu_binding = get_least_used_cpu_on_node(dd->numa_node); + dd->work[2].cpu_binding = get_least_used_cpu_on_node(dd->numa_node); + dd->work[3].cpu_binding = dd->work[0].cpu_binding; + dd->work[4].cpu_binding = dd->work[1].cpu_binding; + dd->work[5].cpu_binding = dd->work[2].cpu_binding; + dd->work[6].cpu_binding = dd->work[2].cpu_binding; + dd->work[7].cpu_binding = dd->work[1].cpu_binding; + + /* Log the bindings */ + for_each_present_cpu(cpu) { + memset(cpu_list, 0, sizeof(cpu_list)); + for (i = 0, j = 0; i < MTIP_MAX_SLOT_GROUPS; i++) { + if (dd->work[i].cpu_binding == cpu) { + snprintf(&cpu_list[j], 256 - j, "%d ", i); + j = strlen(cpu_list); + } + } + if (j) + dev_info(&pdev->dev, "CPU %d: WQs %s\n", cpu, cpu_list); + } + + INIT_WORK(&dd->work[0].work, mtip_workq_sdbf0); + INIT_WORK(&dd->work[1].work, mtip_workq_sdbf1); + INIT_WORK(&dd->work[2].work, mtip_workq_sdbf2); + INIT_WORK(&dd->work[3].work, mtip_workq_sdbf3); + INIT_WORK(&dd->work[4].work, mtip_workq_sdbf4); + INIT_WORK(&dd->work[5].work, mtip_workq_sdbf5); + INIT_WORK(&dd->work[6].work, mtip_workq_sdbf6); + INIT_WORK(&dd->work[7].work, mtip_workq_sdbf7); + + pci_set_master(pdev); if (pci_enable_msi(pdev)) { dev_warn(&pdev->dev, "Unable to enable MSI interrupt.\n"); goto block_initialize_err; } - /* Copy the info we may need later into the private data structure. */ - dd->major = mtip_major; - dd->instance = instance; - dd->pdev = pdev; - /* Initialize the block layer. */ rv = mtip_block_initialize(dd); if (rv < 0) { @@ -4048,7 +4307,13 @@ static int mtip_pci_probe(struct pci_dev *pdev, block_initialize_err: pci_disable_msi(pdev); - + if (dd->isr_workq) { + flush_workqueue(dd->isr_workq); + destroy_workqueue(dd->isr_workq); + drop_cpu(dd->work[0].cpu_binding); + drop_cpu(dd->work[1].cpu_binding); + drop_cpu(dd->work[2].cpu_binding); + } setmask_err: pcim_iounmap_regions(pdev, 1 << MTIP_ABAR); @@ -4089,6 +4354,14 @@ static void mtip_pci_remove(struct pci_dev *pdev) /* Clean up the block layer. */ mtip_block_remove(dd); + if (dd->isr_workq) { + flush_workqueue(dd->isr_workq); + destroy_workqueue(dd->isr_workq); + drop_cpu(dd->work[0].cpu_binding); + drop_cpu(dd->work[1].cpu_binding); + drop_cpu(dd->work[2].cpu_binding); + } + pci_disable_msi(pdev); kfree(dd); diff --git a/drivers/block/mtip32xx/mtip32xx.h b/drivers/block/mtip32xx/mtip32xx.h index b1742640556..3bffff5f670 100644 --- a/drivers/block/mtip32xx/mtip32xx.h +++ b/drivers/block/mtip32xx/mtip32xx.h @@ -164,6 +164,35 @@ struct smart_attr { u8 res[3]; } __packed; +struct mtip_work { + struct work_struct work; + void *port; + int cpu_binding; + u32 completed; +} ____cacheline_aligned_in_smp; + +#define DEFINE_HANDLER(group) \ + void mtip_workq_sdbf##group(struct work_struct *work) \ + { \ + struct mtip_work *w = (struct mtip_work *) work; \ + mtip_workq_sdbfx(w->port, group, w->completed); \ + } + +#define MTIP_TRIM_TIMEOUT_MS 240000 +#define MTIP_MA |