From f30e6d3e419bfb5540fa82ba7eca01d578556e6b Mon Sep 17 00:00:00 2001 From: Stefan Richter Date: Fri, 22 Apr 2011 15:13:54 +0200 Subject: firewire: octlet AT payloads can be stack-allocated We do not need slab allocations anymore in order to satisfy streaming DMA mapping constraints, thanks to commit da28947e7e36 "firewire: ohci: avoid separate DMA mapping for small AT payloads". (Besides, the slab-allocated buffers that firewire-core, firewire-sbp2, and firedtv used to provide for 8-byte write and lock requests were still not fully portable since they crossed cacheline boundaries or shared a cacheline with unrelated CPU-accessed data. snd-firewire-lib got this aspect right by using an extra kmalloc/ kfree just for the 8-byte transaction buffer.) This change replaces kmalloc'ed lock transaction scratch buffers in firewire-core, firedtv, and snd-firewire-lib by local stack allocations. Perhaps the most notable result of the change is simpler locking because there is no need to serialize usages of preallocated per-device buffers anymore. Also, allocations and deallocations are simpler. Signed-off-by: Stefan Richter Acked-by: Clemens Ladisch --- include/linux/firewire.h | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'include/linux/firewire.h') diff --git a/include/linux/firewire.h b/include/linux/firewire.h index c64f3680d4f1..de90e1ff8488 100644 --- a/include/linux/firewire.h +++ b/include/linux/firewire.h @@ -125,7 +125,6 @@ struct fw_card { struct delayed_work bm_work; /* bus manager job */ int bm_retries; int bm_generation; - __be32 bm_transaction_data[2]; int bm_node_id; bool bm_abdicate; @@ -447,6 +446,6 @@ int fw_iso_context_stop(struct fw_iso_context *ctx); void fw_iso_context_destroy(struct fw_iso_context *ctx); void fw_iso_resource_manage(struct fw_card *card, int generation, u64 channels_mask, int *channel, int *bandwidth, - bool allocate, __be32 buffer[2]); + bool allocate); #endif /* _LINUX_FIREWIRE_H */ -- cgit v1.2.3 From 13882a82ee1646336c3996c93b4a560a55d2a419 Mon Sep 17 00:00:00 2001 From: Clemens Ladisch Date: Mon, 2 May 2011 09:33:56 +0200 Subject: firewire: optimize iso queueing by setting wake only after the last packet When queueing iso packets, the run time is dominated by the two MMIO accesses that set the DMA context's wake bit. Because most drivers submit packets in batches, we can save much time by removing all but the last wakeup. The internal kernel API is changed to require a call to fw_iso_context_queue_flush() after a batch of queued packets. The user space API does not change, so one call to FW_CDEV_IOC_QUEUE_ISO must specify multiple packets to take advantage of this optimization. In my measurements, this patch reduces the time needed to queue fifty skip packets from userspace to one sixth on a 2.5 GHz CPU, or to one third at 800 MHz. Signed-off-by: Clemens Ladisch Signed-off-by: Stefan Richter --- drivers/firewire/core-card.c | 5 +++++ drivers/firewire/core-cdev.c | 1 + drivers/firewire/core-iso.c | 6 ++++++ drivers/firewire/core.h | 2 ++ drivers/firewire/net.c | 4 +++- drivers/firewire/ohci.c | 19 +++++++++++++++---- drivers/media/dvb/firewire/firedtv-fw.c | 1 + include/linux/firewire.h | 1 + sound/firewire/amdtp.c | 1 + 9 files changed, 35 insertions(+), 5 deletions(-) (limited to 'include/linux/firewire.h') diff --git a/drivers/firewire/core-card.c b/drivers/firewire/core-card.c index e119f1e6ba47..f05fc7bfceeb 100644 --- a/drivers/firewire/core-card.c +++ b/drivers/firewire/core-card.c @@ -630,6 +630,10 @@ static int dummy_queue_iso(struct fw_iso_context *ctx, struct fw_iso_packet *p, return -ENODEV; } +static void dummy_flush_queue_iso(struct fw_iso_context *ctx) +{ +} + static const struct fw_card_driver dummy_driver_template = { .read_phy_reg = dummy_read_phy_reg, .update_phy_reg = dummy_update_phy_reg, @@ -641,6 +645,7 @@ static const struct fw_card_driver dummy_driver_template = { .start_iso = dummy_start_iso, .set_iso_channels = dummy_set_iso_channels, .queue_iso = dummy_queue_iso, + .flush_queue_iso = dummy_flush_queue_iso, }; void fw_card_release(struct kref *kref) diff --git a/drivers/firewire/core-cdev.c b/drivers/firewire/core-cdev.c index 2a3f1c4d6906..64768c2194f1 100644 --- a/drivers/firewire/core-cdev.c +++ b/drivers/firewire/core-cdev.c @@ -1107,6 +1107,7 @@ static int ioctl_queue_iso(struct client *client, union ioctl_arg *arg) payload += u.packet.payload_length; count++; } + fw_iso_context_queue_flush(ctx); a->size -= uptr_to_u64(p) - a->packets; a->packets = uptr_to_u64(p); diff --git a/drivers/firewire/core-iso.c b/drivers/firewire/core-iso.c index f872ede5af37..57c3973093ad 100644 --- a/drivers/firewire/core-iso.c +++ b/drivers/firewire/core-iso.c @@ -185,6 +185,12 @@ int fw_iso_context_queue(struct fw_iso_context *ctx, } EXPORT_SYMBOL(fw_iso_context_queue); +void fw_iso_context_queue_flush(struct fw_iso_context *ctx) +{ + ctx->card->driver->flush_queue_iso(ctx); +} +EXPORT_SYMBOL(fw_iso_context_queue_flush); + int fw_iso_context_stop(struct fw_iso_context *ctx) { return ctx->card->driver->stop_iso(ctx); diff --git a/drivers/firewire/core.h b/drivers/firewire/core.h index 25e729cde2f7..0fe4e4e6eda7 100644 --- a/drivers/firewire/core.h +++ b/drivers/firewire/core.h @@ -97,6 +97,8 @@ struct fw_card_driver { struct fw_iso_buffer *buffer, unsigned long payload); + void (*flush_queue_iso)(struct fw_iso_context *ctx); + int (*stop_iso)(struct fw_iso_context *ctx); }; diff --git a/drivers/firewire/net.c b/drivers/firewire/net.c index 3f04dd3681cf..b9762d07198d 100644 --- a/drivers/firewire/net.c +++ b/drivers/firewire/net.c @@ -881,7 +881,9 @@ static void fwnet_receive_broadcast(struct fw_iso_context *context, spin_unlock_irqrestore(&dev->lock, flags); - if (retval < 0) + if (retval >= 0) + fw_iso_context_queue_flush(dev->broadcast_rcv_context); + else fw_error("requeue failed\n"); } diff --git a/drivers/firewire/ohci.c b/drivers/firewire/ohci.c index f9f55703375e..438e6c831170 100644 --- a/drivers/firewire/ohci.c +++ b/drivers/firewire/ohci.c @@ -1192,9 +1192,6 @@ static void context_append(struct context *ctx, wmb(); /* finish init of new descriptors before branch_address update */ ctx->prev->branch_address = cpu_to_le32(d_bus | z); ctx->prev = find_branch_descriptor(d, z); - - reg_write(ctx->ohci, CONTROL_SET(ctx->regs), CONTEXT_WAKE); - flush_writes(ctx->ohci); } static void context_stop(struct context *ctx) @@ -1348,8 +1345,12 @@ static int at_context_queue_packet(struct context *ctx, context_append(ctx, d, z, 4 - z); - if (!ctx->running) + if (ctx->running) { + reg_write(ohci, CONTROL_SET(ctx->regs), CONTEXT_WAKE); + flush_writes(ohci); + } else { context_run(ctx, 0); + } return 0; } @@ -3121,6 +3122,15 @@ static int ohci_queue_iso(struct fw_iso_context *base, return ret; } +static void ohci_flush_queue_iso(struct fw_iso_context *base) +{ + struct context *ctx = + &container_of(base, struct iso_context, base)->context; + + reg_write(ctx->ohci, CONTROL_SET(ctx->regs), CONTEXT_WAKE); + flush_writes(ctx->ohci); +} + static const struct fw_card_driver ohci_driver = { .enable = ohci_enable, .read_phy_reg = ohci_read_phy_reg, @@ -3137,6 +3147,7 @@ static const struct fw_card_driver ohci_driver = { .free_iso_context = ohci_free_iso_context, .set_iso_channels = ohci_set_iso_channels, .queue_iso = ohci_queue_iso, + .flush_queue_iso = ohci_flush_queue_iso, .start_iso = ohci_start_iso, .stop_iso = ohci_stop_iso, }; diff --git a/drivers/media/dvb/firewire/firedtv-fw.c b/drivers/media/dvb/firewire/firedtv-fw.c index 8022b743af91..864b6274c729 100644 --- a/drivers/media/dvb/firewire/firedtv-fw.c +++ b/drivers/media/dvb/firewire/firedtv-fw.c @@ -125,6 +125,7 @@ static void handle_iso(struct fw_iso_context *context, u32 cycle, i = (i + 1) & (N_PACKETS - 1); } + fw_iso_context_queue_flush(ctx->context); ctx->current_packet = i; } diff --git a/include/linux/firewire.h b/include/linux/firewire.h index de90e1ff8488..c0fb405bb435 100644 --- a/include/linux/firewire.h +++ b/include/linux/firewire.h @@ -440,6 +440,7 @@ int fw_iso_context_queue(struct fw_iso_context *ctx, struct fw_iso_packet *packet, struct fw_iso_buffer *buffer, unsigned long payload); +void fw_iso_context_queue_flush(struct fw_iso_context *ctx); int fw_iso_context_start(struct fw_iso_context *ctx, int cycle, int sync, int tags); int fw_iso_context_stop(struct fw_iso_context *ctx); diff --git a/sound/firewire/amdtp.c b/sound/firewire/amdtp.c index b18140ff2b93..87657dd7714c 100644 --- a/sound/firewire/amdtp.c +++ b/sound/firewire/amdtp.c @@ -396,6 +396,7 @@ static void out_packet_callback(struct fw_iso_context *context, u32 cycle, for (i = 0; i < packets; ++i) queue_out_packet(s, ++cycle); + fw_iso_context_queue_flush(s->context); } static int queue_initial_skip_packets(struct amdtp_out_stream *s) -- cgit v1.2.3 From 105e53f863c04e1d9e5bb34bf753c9fdbce6a60c Mon Sep 17 00:00:00 2001 From: Stefan Richter Date: Sun, 1 May 2011 20:50:31 +0200 Subject: firewire: sbp2: parallelize login, reconnect, logout The struct sbp2_logical_unit.work items can all be executed in parallel but are not reentrant. Furthermore, reconnect or re-login work must be executed in a WQ_MEM_RECLAIM workqueue. Hence replace the old single-threaded firewire-sbp2 workqueue by a concurrency-managed but non-reentrant workqueue with rescuer. firewire-core already maintains one, hence use this one. In earlier versions of this change, I observed occasional failures of parallel INQUIRY to an Initio INIC-2430 FireWire 800 to dual IDE bridge. More testing indicates that parallel INQUIRY is not actually a problem, but too quick successions of logout and login + INQUIRY, e.g. a quick sequence of cable plugout and plugin, can result in failed INQUIRY. This does not seem to be something that should or could be addressed by serialization. Another dual-LU device to which I currently have access to, an OXUF924DSB FireWire 800 to dual SATA bridge with firmware from MacPower, has been successfully tested with this too. This change is beneficial to environments with two or more FireWire storage devices, especially if they are located on the same bus. Management tasks that should be performed as soon and as quickly as possible, especially reconnect, are no longer held up by tasks on other devices that may take a long time, especially login with INQUIRY and sd or sr driver probe. Signed-off-by: Stefan Richter --- drivers/firewire/core-card.c | 4 ++-- drivers/firewire/core-cdev.c | 2 +- drivers/firewire/core-device.c | 5 +++-- drivers/firewire/core-transaction.c | 12 ++++++------ drivers/firewire/core.h | 2 -- drivers/firewire/sbp2.c | 9 +-------- include/linux/firewire.h | 2 ++ 7 files changed, 15 insertions(+), 21 deletions(-) (limited to 'include/linux/firewire.h') diff --git a/drivers/firewire/core-card.c b/drivers/firewire/core-card.c index bb8c4d22b03e..29d2423fae6d 100644 --- a/drivers/firewire/core-card.c +++ b/drivers/firewire/core-card.c @@ -228,7 +228,7 @@ void fw_schedule_bus_reset(struct fw_card *card, bool delayed, bool short_reset) /* Use an arbitrary short delay to combine multiple reset requests. */ fw_card_get(card); - if (!queue_delayed_work(fw_wq, &card->br_work, + if (!queue_delayed_work(fw_workqueue, &card->br_work, delayed ? DIV_ROUND_UP(HZ, 100) : 0)) fw_card_put(card); } @@ -241,7 +241,7 @@ static void br_work(struct work_struct *work) /* Delay for 2s after last reset per IEEE 1394 clause 8.2.1. */ if (card->reset_jiffies != 0 && time_before64(get_jiffies_64(), card->reset_jiffies + 2 * HZ)) { - if (!queue_delayed_work(fw_wq, &card->br_work, 2 * HZ)) + if (!queue_delayed_work(fw_workqueue, &card->br_work, 2 * HZ)) fw_card_put(card); return; } diff --git a/drivers/firewire/core-cdev.c b/drivers/firewire/core-cdev.c index aa1131d26e30..b1c11775839c 100644 --- a/drivers/firewire/core-cdev.c +++ b/drivers/firewire/core-cdev.c @@ -149,7 +149,7 @@ static void release_iso_resource(struct client *, struct client_resource *); static void schedule_iso_resource(struct iso_resource *r, unsigned long delay) { client_get(r->client); - if (!queue_delayed_work(fw_wq, &r->work, delay)) + if (!queue_delayed_work(fw_workqueue, &r->work, delay)) client_put(r->client); } diff --git a/drivers/firewire/core-device.c b/drivers/firewire/core-device.c index ef900d923f15..95a471401892 100644 --- a/drivers/firewire/core-device.c +++ b/drivers/firewire/core-device.c @@ -725,12 +725,13 @@ struct fw_device *fw_device_get_by_devt(dev_t devt) return device; } -struct workqueue_struct *fw_wq; +struct workqueue_struct *fw_workqueue; +EXPORT_SYMBOL(fw_workqueue); static void fw_schedule_device_work(struct fw_device *device, unsigned long delay) { - queue_delayed_work(fw_wq, &device->work, delay); + queue_delayed_work(fw_workqueue, &device->work, delay); } /* diff --git a/drivers/firewire/core-transaction.c b/drivers/firewire/core-transaction.c index d4c28a217b2c..334b82a3542c 100644 --- a/drivers/firewire/core-transaction.c +++ b/drivers/firewire/core-transaction.c @@ -1214,21 +1214,21 @@ static int __init fw_core_init(void) { int ret; - fw_wq = alloc_workqueue(KBUILD_MODNAME, - WQ_NON_REENTRANT | WQ_MEM_RECLAIM, 0); - if (!fw_wq) + fw_workqueue = alloc_workqueue("firewire", + WQ_NON_REENTRANT | WQ_MEM_RECLAIM, 0); + if (!fw_workqueue) return -ENOMEM; ret = bus_register(&fw_bus_type); if (ret < 0) { - destroy_workqueue(fw_wq); + destroy_workqueue(fw_workqueue); return ret; } fw_cdev_major = register_chrdev(0, "firewire", &fw_device_ops); if (fw_cdev_major < 0) { bus_unregister(&fw_bus_type); - destroy_workqueue(fw_wq); + destroy_workqueue(fw_workqueue); return fw_cdev_major; } @@ -1244,7 +1244,7 @@ static void __exit fw_core_cleanup(void) { unregister_chrdev(fw_cdev_major, "firewire"); bus_unregister(&fw_bus_type); - destroy_workqueue(fw_wq); + destroy_workqueue(fw_workqueue); idr_destroy(&fw_device_idr); } diff --git a/drivers/firewire/core.h b/drivers/firewire/core.h index 00ea7730c6a7..0fe4e4e6eda7 100644 --- a/drivers/firewire/core.h +++ b/drivers/firewire/core.h @@ -140,8 +140,6 @@ void fw_cdev_handle_phy_packet(struct fw_card *card, struct fw_packet *p); extern struct rw_semaphore fw_device_rwsem; extern struct idr fw_device_idr; extern int fw_cdev_major; -struct workqueue_struct; -extern struct workqueue_struct *fw_wq; struct fw_device *fw_device_get_by_devt(dev_t devt); int fw_device_set_broadcast_channel(struct device *dev, void *gen); diff --git a/drivers/firewire/sbp2.c b/drivers/firewire/sbp2.c index 2aafc614ae14..41841a3e3f99 100644 --- a/drivers/firewire/sbp2.c +++ b/drivers/firewire/sbp2.c @@ -826,8 +826,6 @@ static void sbp2_target_put(struct sbp2_target *tgt) kref_put(&tgt->kref, sbp2_release_target); } -static struct workqueue_struct *sbp2_wq; - /* * Always get the target's kref when scheduling work on one its units. * Each workqueue job is responsible to call sbp2_target_put() upon return. @@ -835,7 +833,7 @@ static struct workqueue_struct *sbp2_wq; static void sbp2_queue_work(struct sbp2_logical_unit *lu, unsigned long delay) { sbp2_target_get(lu->tgt); - if (!queue_delayed_work(sbp2_wq, &lu->work, delay)) + if (!queue_delayed_work(fw_workqueue, &lu->work, delay)) sbp2_target_put(lu->tgt); } @@ -1645,17 +1643,12 @@ MODULE_ALIAS("sbp2"); static int __init sbp2_init(void) { - sbp2_wq = create_singlethread_workqueue(KBUILD_MODNAME); - if (!sbp2_wq) - return -ENOMEM; - return driver_register(&sbp2_driver.driver); } static void __exit sbp2_cleanup(void) { driver_unregister(&sbp2_driver.driver); - destroy_workqueue(sbp2_wq); } module_init(sbp2_init); diff --git a/include/linux/firewire.h b/include/linux/firewire.h index c0fb405bb435..5e6f42789afe 100644 --- a/include/linux/firewire.h +++ b/include/linux/firewire.h @@ -449,4 +449,6 @@ void fw_iso_resource_manage(struct fw_card *card, int generation, u64 channels_mask, int *channel, int *bandwidth, bool allocate); +extern struct workqueue_struct *fw_workqueue; + #endif /* _LINUX_FIREWIRE_H */ -- cgit v1.2.3