linux/drivers/ide/ide-io.c

910 lines
24 KiB
C
Raw Normal View History

/*
* IDE I/O functions
*
* Basic PIO and command management functionality.
*
* This code was split off from ide.c. See ide.c for history and original
* copyrights.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2, or (at your option) any
* later version.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* For the avoidance of doubt the "preferred form" of this code is one which
* is in an open non patent encumbered format. Where cryptographic key signing
* forms part of the process of creating an executable the information
* including keys needed to generate an equivalently functional executable
* are deemed to be part of the source code.
*/
#include <linux/module.h>
#include <linux/types.h>
#include <linux/string.h>
#include <linux/kernel.h>
#include <linux/timer.h>
#include <linux/mm.h>
#include <linux/interrupt.h>
#include <linux/major.h>
#include <linux/errno.h>
#include <linux/genhd.h>
#include <linux/blkpg.h>
#include <linux/slab.h>
#include <linux/init.h>
#include <linux/pci.h>
#include <linux/delay.h>
#include <linux/ide.h>
#include <linux/completion.h>
#include <linux/reboot.h>
#include <linux/cdrom.h>
#include <linux/seq_file.h>
#include <linux/device.h>
#include <linux/kmod.h>
#include <linux/scatterlist.h>
#include <linux/bitops.h>
#include <asm/byteorder.h>
#include <asm/irq.h>
#include <asm/uaccess.h>
#include <asm/io.h>
int ide_end_rq(ide_drive_t *drive, struct request *rq, int error,
unsigned int nr_bytes)
{
/*
* decide whether to reenable DMA -- 3 is a random magic for now,
* if we DMA timeout more than 3 times, just stay in PIO
*/
if ((drive->dev_flags & IDE_DFLAG_DMA_PIO_RETRY) &&
drive->retry_pio <= 3) {
drive->dev_flags &= ~IDE_DFLAG_DMA_PIO_RETRY;
ide_dma_on(drive);
}
return blk_end_request(rq, error, nr_bytes);
}
EXPORT_SYMBOL_GPL(ide_end_rq);
void ide_complete_cmd(ide_drive_t *drive, struct ide_cmd *cmd, u8 stat, u8 err)
{
const struct ide_tp_ops *tp_ops = drive->hwif->tp_ops;
struct ide_taskfile *tf = &cmd->tf;
struct request *rq = cmd->rq;
u8 tf_cmd = tf->command;
tf->error = err;
tf->status = stat;
if (cmd->ftf_flags & IDE_FTFLAG_IN_DATA) {
u8 data[2];
tp_ops->input_data(drive, cmd, data, 2);
cmd->tf.data = data[0];
cmd->hob.data = data[1];
}
ide_tf_readback(drive, cmd);
if ((cmd->tf_flags & IDE_TFLAG_CUSTOM_HANDLER) &&
tf_cmd == ATA_CMD_IDLEIMMEDIATE) {
if (tf->lbal != 0xc4) {
printk(KERN_ERR "%s: head unload failed!\n",
drive->name);
ide_tf_dump(drive->name, cmd);
} else
drive->dev_flags |= IDE_DFLAG_PARKED;
}
if (rq && rq->cmd_type == REQ_TYPE_ATA_TASKFILE) {
struct ide_cmd *orig_cmd = rq->special;
if (cmd->tf_flags & IDE_TFLAG_DYN)
kfree(orig_cmd);
else
memcpy(orig_cmd, cmd, sizeof(*cmd));
}
}
/* obsolete, blk_rq_bytes() should be used instead */
unsigned int ide_rq_bytes(struct request *rq)
{
if (blk_pc_request(rq))
return blk_rq_bytes(rq);
else
return blk_rq_cur_sectors(rq) << 9;
}
EXPORT_SYMBOL_GPL(ide_rq_bytes);
int ide_complete_rq(ide_drive_t *drive, int error, unsigned int nr_bytes)
{
ide_hwif_t *hwif = drive->hwif;
struct request *rq = hwif->rq;
int rc;
/*
* if failfast is set on a request, override number of sectors
* and complete the whole request right now
*/
if (blk_noretry_request(rq) && error <= 0)
nr_bytes = blk_rq_sectors(rq) << 9;
rc = ide_end_rq(drive, rq, error, nr_bytes);
if (rc == 0)
hwif->rq = NULL;
return rc;
}
EXPORT_SYMBOL(ide_complete_rq);
void ide_kill_rq(ide_drive_t *drive, struct request *rq)
{
u8 drv_req = blk_special_request(rq) && rq->rq_disk;
u8 media = drive->media;
drive->failed_pc = NULL;
if ((media == ide_floppy || media == ide_tape) && drv_req) {
rq->errors = 0;
ide_complete_rq(drive, 0, blk_rq_bytes(rq));
} else {
if (media == ide_tape)
rq->errors = IDE_DRV_ERROR_GENERAL;
else if (blk_fs_request(rq) == 0 && rq->errors == 0)
rq->errors = -EIO;
ide_complete_rq(drive, -EIO, ide_rq_bytes(rq));
}
}
static void ide_tf_set_specify_cmd(ide_drive_t *drive, struct ide_taskfile *tf)
{
tf->nsect = drive->sect;
tf->lbal = drive->sect;
tf->lbam = drive->cyl;
tf->lbah = drive->cyl >> 8;
tf->device = (drive->head - 1) | drive->select;
tf->command = ATA_CMD_INIT_DEV_PARAMS;
}
static void ide_tf_set_restore_cmd(ide_drive_t *drive, struct ide_taskfile *tf)
{
tf->nsect = drive->sect;
tf->command = ATA_CMD_RESTORE;
}
static void ide_tf_set_setmult_cmd(ide_drive_t *drive, struct ide_taskfile *tf)
{
tf->nsect = drive->mult_req;
tf->command = ATA_CMD_SET_MULTI;
}
/**
* do_special - issue some special commands
* @drive: drive the command is for
*
* do_special() is used to issue ATA_CMD_INIT_DEV_PARAMS,
* ATA_CMD_RESTORE and ATA_CMD_SET_MULTI commands to a drive.
*/
static ide_startstop_t do_special(ide_drive_t *drive)
{
struct ide_cmd cmd;
#ifdef DEBUG
printk(KERN_DEBUG "%s: %s: 0x%02x\n", drive->name, __func__,
drive->special_flags);
#endif
if (drive->media != ide_disk) {
drive->special_flags = 0;
drive->mult_req = 0;
return ide_stopped;
}
memset(&cmd, 0, sizeof(cmd));
cmd.protocol = ATA_PROT_NODATA;
if (drive->special_flags & IDE_SFLAG_SET_GEOMETRY) {
drive->special_flags &= ~IDE_SFLAG_SET_GEOMETRY;
ide_tf_set_specify_cmd(drive, &cmd.tf);
} else if (drive->special_flags & IDE_SFLAG_RECALIBRATE) {
drive->special_flags &= ~IDE_SFLAG_RECALIBRATE;
ide_tf_set_restore_cmd(drive, &cmd.tf);
} else if (drive->special_flags & IDE_SFLAG_SET_MULTMODE) {
drive->special_flags &= ~IDE_SFLAG_SET_MULTMODE;
ide_tf_set_setmult_cmd(drive, &cmd.tf);
} else
BUG();
cmd.valid.out.tf = IDE_VALID_OUT_TF | IDE_VALID_DEVICE;
cmd.valid.in.tf = IDE_VALID_IN_TF | IDE_VALID_DEVICE;
cmd.tf_flags = IDE_TFLAG_CUSTOM_HANDLER;
do_rw_taskfile(drive, &cmd);
return ide_started;
}
void ide_map_sg(ide_drive_t *drive, struct ide_cmd *cmd)
{
ide_hwif_t *hwif = drive->hwif;
struct scatterlist *sg = hwif->sg_table;
struct request *rq = cmd->rq;
cmd->sg_nents = blk_rq_map_sg(drive->queue, rq, sg);
}
EXPORT_SYMBOL_GPL(ide_map_sg);
void ide_init_sg_cmd(struct ide_cmd *cmd, unsigned int nr_bytes)
{
cmd->nbytes = cmd->nleft = nr_bytes;
cmd->cursg_ofs = 0;
cmd->cursg = NULL;
}
EXPORT_SYMBOL_GPL(ide_init_sg_cmd);
/**
* execute_drive_command - issue special drive command
* @drive: the drive to issue the command on
* @rq: the request structure holding the command
*
* execute_drive_cmd() issues a special drive command, usually
* initiated by ioctl() from the external hdparm program. The
* command can be a drive command, drive task or taskfile
* operation. Weirdly you can call it with NULL to wait for
* all commands to finish. Don't do this as that is due to change
*/
static ide_startstop_t execute_drive_cmd (ide_drive_t *drive,
struct request *rq)
{
struct ide_cmd *cmd = rq->special;
if (cmd) {
if (cmd->protocol == ATA_PROT_PIO) {
ide_init_sg_cmd(cmd, blk_rq_sectors(rq) << 9);
ide_map_sg(drive, cmd);
}
return do_rw_taskfile(drive, cmd);
}
/*
* NULL is actually a valid way of waiting for
* all current requests to be flushed from the queue.
*/
#ifdef DEBUG
printk("%s: DRIVE_CMD (null)\n", drive->name);
#endif
rq->errors = 0;
ide_complete_rq(drive, 0, blk_rq_bytes(rq));
return ide_stopped;
}
static ide_startstop_t ide_special_rq(ide_drive_t *drive, struct request *rq)
{
ide: Implement disk shock protection support (v4) On user request (through sysfs), the IDLE IMMEDIATE command with UNLOAD FEATURE as specified in ATA-7 is issued to the device and processing of the request queue is stopped thereafter until the specified timeout expires or user space asks to resume normal operation. This is supposed to prevent the heads of a hard drive from accidentally crashing onto the platter when a heavy shock is anticipated (like a falling laptop expected to hit the floor). Port resets are deferred whenever a device on that port is in the parked state. v3: Elias Oltmanns <eo@nebensachen.de> wrote: [...] > >> 1. Make sure that no negative value is being passed to > >> jiffies_to_msecs() in ide_park_show(). > >> 2. Drop the superfluous variable hwif in ide_special_rq(). > >> 3. Skip initialisation of task and tf in ide_special_rq() if we are not > >> handling a (un)park request. > > > > Well, #3 should have been done differently because we donn't want to > > check for REQ_(UN)?PARK_HEADS more often than is necessary. > > While preparing the backport to 2.6.27, it has just occurred to me that > we need to clear the IDE_DFLAG_PARKED flag in ide_disk_pre_reset() > because this flag must not be set after *any* sort of access to the > device. v4: Fix a memory leak due to a missing blk_put_request() in issue_park_cmd(). Additionally, we should plug the queue when enqueueing the unpark request because there is no guarantee that the park timeout has not expired by then. Even though the chance for that to happen is very slim, the request might end up hanging in the queue until the next I/O operation is queued up. While at it, clean up the code a little: - make issue_park_cmd() a function of type void since nobody cares for the return value anyway; - use blk_start_queueing() instead of __blk_run_queue() since we don't have to worry about recursion; - remove a superfluous pointer deference in task_no_data_intr(). Signed-off-by: Elias Oltmanns <eo@nebensachen.de> Cc: Jeff Garzik <jeff@garzik.org>, Cc: Randy Dunlap <randy.dunlap@oracle.com> Cc: Tejun Heo <htejun@gmail.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2008-10-13 19:39:50 +00:00
u8 cmd = rq->cmd[0];
switch (cmd) {
case REQ_PARK_HEADS:
case REQ_UNPARK_HEADS:
return ide_do_park_unpark(drive, rq);
case REQ_DEVSET_EXEC:
return ide_do_devset(drive, rq);
case REQ_DRIVE_RESET:
return ide_do_reset(drive);
default:
BUG();
}
}
/**
* start_request - start of I/O and command issuing for IDE
*
* start_request() initiates handling of a new I/O request. It
* accepts commands and I/O (read/write) requests.
*
* FIXME: this function needs a rename
*/
static ide_startstop_t start_request (ide_drive_t *drive, struct request *rq)
{
ide_startstop_t startstop;
BUG_ON(!blk_rq_started(rq));
#ifdef DEBUG
printk("%s: start_request: current=0x%08lx\n",
drive->hwif->name, (unsigned long) rq);
#endif
/* bail early if we've exceeded max_failures */
if (drive->max_failures && (drive->failures > drive->max_failures)) {
ide-io: set REQ_FAILED when drive is dead Currently it's possible to ide-cd to set an incorrect blocksize by reading garbage if the drive is dead: ide_cd_probe() -> cdrom_read_toc() -> cdrom_read_capacity() -> cdrom_queue_packet_command() -> ide_do_drive_cmd() -> ide_do_request() -> start_request() on start_request(): /* bail early if we've exceeded max_failures */ if (drive->max_failures && (drive->failures > drive->max_failures)) { goto kill_rq; } (...) kill_rq: ide_kill_rq(drive, rq); return ide_stopped; ide_kill_rq() and the next calls won't set REQ_FAILED on rq->cmd_flags and thus cdrom_queue_packet_command() won't return an error. then: stat = cdrom_queue_packet_command(drive, &req); if (stat == 0) { *capacity = 1 + be32_to_cpu(capbuf.lba); *sectors_per_frame = be32_to_cpu(capbuf.blocklen) >> SECTOR_BITS; } cdrom_read_capacity() ends believing capbuf is valid but in fact it's just uninitialized data. back to cdrom_read_toc(): /* Try to get the total cdrom capacity and sector size. */ stat = cdrom_read_capacity(drive, &toc->capacity, &sectors_per_frame, sense); if (stat) toc->capacity = 0x1fffff; set_capacity(info->disk, toc->capacity * sectors_per_frame); /* Save a private copy of te TOC capacity for error handling */ drive->probed_capacity = toc->capacity * sectors_per_frame; blk_queue_hardsect_size(drive->queue, sectors_per_frame << SECTOR_BITS); that will set drive->queue->hardsect_size to be the random value. hardsect_size is used to calculate inode->i_blkbits. later on, on a read path: void create_empty_buffers(struct page *page, unsigned long blocksize, unsigned long b_state) { struct buffer_head *bh, *head, *tail; head = alloc_page_buffers(page, blocksize, 1); bh = head; do { bh->b_state |= b_state; tail = bh; bh = bh->b_this_page; } while (bh); tail->b_this_page = head; alloc_page_buffers() will return NULL if blocksize > 4096. blocksize is calculed based on inode->i_blkbits. that will trigger a null dereference on create_empty_buffers(). Signed-off-by: Aristeu Rozanski <arozansk@redhat.com> Cc: Borislav Petkov <bbpetkov@yahoo.de> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2008-01-25 21:17:04 +00:00
rq->cmd_flags |= REQ_FAILED;
goto kill_rq;
}
if (blk_pm_request(rq))
ide_check_pm_state(drive, rq);
drive->hwif->tp_ops->dev_select(drive);
if (ide_wait_stat(&startstop, drive, drive->ready_stat,
ATA_BUSY | ATA_DRQ, WAIT_READY)) {
printk(KERN_ERR "%s: drive not ready for command\n", drive->name);
return startstop;
}
if (drive->special_flags == 0) {
struct ide_driver *drv;
/*
* We reset the drive so we need to issue a SETFEATURES.
* Do it _after_ do_special() restored device parameters.
*/
if (drive->current_speed == 0xff)
ide_config_drive_speed(drive, drive->desired_speed);
if (rq->cmd_type == REQ_TYPE_ATA_TASKFILE)
return execute_drive_cmd(drive, rq);
else if (blk_pm_request(rq)) {
struct request_pm_state *pm = rq->special;
#ifdef DEBUG_PM
printk("%s: start_power_step(step: %d)\n",
drive->name, pm->pm_step);
#endif
startstop = ide_start_power_step(drive, rq);
if (startstop == ide_stopped &&
pm->pm_step == IDE_PM_COMPLETED)
ide_complete_pm_rq(drive, rq);
return startstop;
} else if (!rq->rq_disk && blk_special_request(rq))
/*
* TODO: Once all ULDs have been modified to
* check for specific op codes rather than
* blindly accepting any special request, the
* check for ->rq_disk above may be replaced
* by a more suitable mechanism or even
* dropped entirely.
*/
return ide_special_rq(drive, rq);
drv = *(struct ide_driver **)rq->rq_disk->private_data;
return drv->do_request(drive, rq, blk_rq_pos(rq));
}
return do_special(drive);
kill_rq:
ide_kill_rq(drive, rq);
return ide_stopped;
}
/**
* ide_stall_queue - pause an IDE device
* @drive: drive to stall
* @timeout: time to stall for (jiffies)
*
* ide_stall_queue() can be used by a drive to give excess bandwidth back
* to the port by sleeping for timeout jiffies.
*/
void ide_stall_queue (ide_drive_t *drive, unsigned long timeout)
{
if (timeout > WAIT_WORSTCASE)
timeout = WAIT_WORSTCASE;
drive->sleep = timeout + jiffies;
drive->dev_flags |= IDE_DFLAG_SLEEPING;
}
EXPORT_SYMBOL(ide_stall_queue);
static inline int ide_lock_port(ide_hwif_t *hwif)
{
if (hwif->busy)
return 1;
hwif->busy = 1;
return 0;
}
static inline void ide_unlock_port(ide_hwif_t *hwif)
{
hwif->busy = 0;
}
static inline int ide_lock_host(struct ide_host *host, ide_hwif_t *hwif)
{
int rc = 0;
if (host->host_flags & IDE_HFLAG_SERIALIZE) {
rc = test_and_set_bit_lock(IDE_HOST_BUSY, &host->host_busy);
if (rc == 0) {
if (host->get_lock)
host->get_lock(ide_intr, hwif);
}
}
return rc;
}
static inline void ide_unlock_host(struct ide_host *host)
{
if (host->host_flags & IDE_HFLAG_SERIALIZE) {
if (host->release_lock)
host->release_lock();
clear_bit_unlock(IDE_HOST_BUSY, &host->host_busy);
}
}
/*
* Issue a new request to a device.
*/
ide: don't execute the next queued command from the hard-IRQ context (v2) * Tell the block layer that we are not done handling requests by using blk_plug_device() in ide_do_request() (request handling function) and ide_timer_expiry() (timeout handler) if the queue is not empty. * Remove optimization which directly calls ide_do_request() for the next queued command from the ide_intr() (IRQ handler) and ide_timer_expiry(). * Remove no longer needed IRQ masking from ide_do_request() - in case of IDE ports needing serialization disable_irq_nosync()/enable_irq() was used for the (possibly shared) IRQ of the other IDE port. * Put the misplaced comment in the right place in ide_do_request(). * Drop no longer needed 'int masked_irq' argument from ide_do_request(). * Merge ide_do_request() into do_ide_request(). * Remove no longer needed IDE_NO_IRQ define. While at it: * Don't use HWGROUP() macro in do_ide_request(). * Use __func__ in ide_intr(). This patch reduces IRQ hadling latency for IDE and improves the system-wide handling of shared IRQs (which should result in more timeout resistant and stable IDE systems). It also makes it possible to do some further changes later (i.e. replace some busy-waiting delays with sleeping equivalents). v2: Changes per review from Elias Oltmanns: - fix wrong goto statement in 'if (startstop == ide_stopped)' block - use spin_unlock_irq() - don't use obsolete HWIF() macro Cc: Elias Oltmanns <eo@nebensachen.de> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2009-01-02 15:12:48 +00:00
void do_ide_request(struct request_queue *q)
{
ide_drive_t *drive = q->queuedata;
ide_hwif_t *hwif = drive->hwif;
struct ide_host *host = hwif->host;
struct request *rq = NULL;
ide_startstop_t startstop;
/*
* drive is doing pre-flush, ordered write, post-flush sequence. even
* though that is 3 requests, it must be seen as a single transaction.
* we must not preempt this drive until that is complete
*/
if (blk_queue_flushing(q))
/*
* small race where queue could get replugged during
* the 3-request flush cycle, just yank the plug since
* we want it to finish asap
*/
blk_remove_plug(q);
spin_unlock_irq(q->queue_lock);
ide-cd,atapi: use bio for internal commands Impact: unify request data buffer handling rq->data is used mostly to pass kernel buffer through request queue without using bio. There are only a couple of places which still do this in kernel and converting to bio isn't difficult. This patch converts ide-cd and atapi to use bio instead of rq->data for request sense and internal pc commands. With previous change to unify sense request handling, this is relatively easily achieved by adding blk_rq_map_kern() during sense_rq prep and PC issue. If blk_rq_map_kern() fails for sense, the error is deferred till sense issue and aborts the failed command which triggered the sense. Note that this is a slim possibility as sense prep is done on each command issue, so for the above condition to actually trigger, all preps since the last sense issue till the issue of the request which would require a sense should fail. * do_request functions might sleep now. This should be okay as ide request_fn - do_ide_request() - is invoked only from make_request and plug work. Make sure this is the case by adding might_sleep() to do_ide_request(). * Functions which access the read sense data before the sense request is complete now should access bio_data(sense_rq->bio) as the sense buffer might have been copied during blk_rq_map_kern(). * ide-tape updated to map sg. * cdrom_do_block_pc() now doesn't have to deal with REQ_TYPE_ATA_PC special case. Simplified. * tp_ops->output/input_data path dropped from ide_pc_intr(). Signed-off-by: Tejun Heo <tj@kernel.org>
2009-04-18 22:00:42 +00:00
/* HLD do_request() callback might sleep, make sure it's okay */
might_sleep();
if (ide_lock_host(host, hwif))
goto plug_device_2;
spin_lock_irq(&hwif->lock);
ide: don't execute the next queued command from the hard-IRQ context (v2) * Tell the block layer that we are not done handling requests by using blk_plug_device() in ide_do_request() (request handling function) and ide_timer_expiry() (timeout handler) if the queue is not empty. * Remove optimization which directly calls ide_do_request() for the next queued command from the ide_intr() (IRQ handler) and ide_timer_expiry(). * Remove no longer needed IRQ masking from ide_do_request() - in case of IDE ports needing serialization disable_irq_nosync()/enable_irq() was used for the (possibly shared) IRQ of the other IDE port. * Put the misplaced comment in the right place in ide_do_request(). * Drop no longer needed 'int masked_irq' argument from ide_do_request(). * Merge ide_do_request() into do_ide_request(). * Remove no longer needed IDE_NO_IRQ define. While at it: * Don't use HWGROUP() macro in do_ide_request(). * Use __func__ in ide_intr(). This patch reduces IRQ hadling latency for IDE and improves the system-wide handling of shared IRQs (which should result in more timeout resistant and stable IDE systems). It also makes it possible to do some further changes later (i.e. replace some busy-waiting delays with sleeping equivalents). v2: Changes per review from Elias Oltmanns: - fix wrong goto statement in 'if (startstop == ide_stopped)' block - use spin_unlock_irq() - don't use obsolete HWIF() macro Cc: Elias Oltmanns <eo@nebensachen.de> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2009-01-02 15:12:48 +00:00
if (!ide_lock_port(hwif)) {
ide_hwif_t *prev_port;
WARN_ON_ONCE(hwif->rq);
repeat:
prev_port = hwif->host->cur_port;
if (drive->dev_flags & IDE_DFLAG_SLEEPING &&
time_after(drive->sleep, jiffies)) {
ide_unlock_port(hwif);
goto plug_device;
}
ide: don't execute the next queued command from the hard-IRQ context (v2) * Tell the block layer that we are not done handling requests by using blk_plug_device() in ide_do_request() (request handling function) and ide_timer_expiry() (timeout handler) if the queue is not empty. * Remove optimization which directly calls ide_do_request() for the next queued command from the ide_intr() (IRQ handler) and ide_timer_expiry(). * Remove no longer needed IRQ masking from ide_do_request() - in case of IDE ports needing serialization disable_irq_nosync()/enable_irq() was used for the (possibly shared) IRQ of the other IDE port. * Put the misplaced comment in the right place in ide_do_request(). * Drop no longer needed 'int masked_irq' argument from ide_do_request(). * Merge ide_do_request() into do_ide_request(). * Remove no longer needed IDE_NO_IRQ define. While at it: * Don't use HWGROUP() macro in do_ide_request(). * Use __func__ in ide_intr(). This patch reduces IRQ hadling latency for IDE and improves the system-wide handling of shared IRQs (which should result in more timeout resistant and stable IDE systems). It also makes it possible to do some further changes later (i.e. replace some busy-waiting delays with sleeping equivalents). v2: Changes per review from Elias Oltmanns: - fix wrong goto statement in 'if (startstop == ide_stopped)' block - use spin_unlock_irq() - don't use obsolete HWIF() macro Cc: Elias Oltmanns <eo@nebensachen.de> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2009-01-02 15:12:48 +00:00
if ((hwif->host->host_flags & IDE_HFLAG_SERIALIZE) &&
hwif != prev_port) {
ide_drive_t *cur_dev =
prev_port ? prev_port->cur_dev : NULL;
/*
* set nIEN for previous port, drives in the
* quirk list may not like intr setups/cleanups
*/
if (cur_dev &&
(cur_dev->dev_flags & IDE_DFLAG_NIEN_QUIRK) == 0)
prev_port->tp_ops->write_devctl(prev_port,
ATA_NIEN |
ATA_DEVCTL_OBS);
hwif->host->cur_port = hwif;
}
hwif->cur_dev = drive;
ide: Implement disk shock protection support (v4) On user request (through sysfs), the IDLE IMMEDIATE command with UNLOAD FEATURE as specified in ATA-7 is issued to the device and processing of the request queue is stopped thereafter until the specified timeout expires or user space asks to resume normal operation. This is supposed to prevent the heads of a hard drive from accidentally crashing onto the platter when a heavy shock is anticipated (like a falling laptop expected to hit the floor). Port resets are deferred whenever a device on that port is in the parked state. v3: Elias Oltmanns <eo@nebensachen.de> wrote: [...] > >> 1. Make sure that no negative value is being passed to > >> jiffies_to_msecs() in ide_park_show(). > >> 2. Drop the superfluous variable hwif in ide_special_rq(). > >> 3. Skip initialisation of task and tf in ide_special_rq() if we are not > >> handling a (un)park request. > > > > Well, #3 should have been done differently because we donn't want to > > check for REQ_(UN)?PARK_HEADS more often than is necessary. > > While preparing the backport to 2.6.27, it has just occurred to me that > we need to clear the IDE_DFLAG_PARKED flag in ide_disk_pre_reset() > because this flag must not be set after *any* sort of access to the > device. v4: Fix a memory leak due to a missing blk_put_request() in issue_park_cmd(). Additionally, we should plug the queue when enqueueing the unpark request because there is no guarantee that the park timeout has not expired by then. Even though the chance for that to happen is very slim, the request might end up hanging in the queue until the next I/O operation is queued up. While at it, clean up the code a little: - make issue_park_cmd() a function of type void since nobody cares for the return value anyway; - use blk_start_queueing() instead of __blk_run_queue() since we don't have to worry about recursion; - remove a superfluous pointer deference in task_no_data_intr(). Signed-off-by: Elias Oltmanns <eo@nebensachen.de> Cc: Jeff Garzik <jeff@garzik.org>, Cc: Randy Dunlap <randy.dunlap@oracle.com> Cc: Tejun Heo <htejun@gmail.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2008-10-13 19:39:50 +00:00
drive->dev_flags &= ~(IDE_DFLAG_SLEEPING | IDE_DFLAG_PARKED);
spin_unlock_irq(&hwif->lock);
spin_lock_irq(q->queue_lock);
/*
* we know that the queue isn't empty, but this can happen
* if the q->prep_rq_fn() decides to kill a request
*/
block: implement and enforce request peek/start/fetch Till now block layer allowed two separate modes of request execution. A request is always acquired from the request queue via elv_next_request(). After that, drivers are free to either dequeue it or process it without dequeueing. Dequeue allows elv_next_request() to return the next request so that multiple requests can be in flight. Executing requests without dequeueing has its merits mostly in allowing drivers for simpler devices which can't do sg to deal with segments only without considering request boundary. However, the benefit this brings is dubious and declining while the cost of the API ambiguity is increasing. Segment based drivers are usually for very old or limited devices and as converting to dequeueing model isn't difficult, it doesn't justify the API overhead it puts on block layer and its more modern users. Previous patches converted all block low level drivers to dequeueing model. This patch completes the API transition by... * renaming elv_next_request() to blk_peek_request() * renaming blkdev_dequeue_request() to blk_start_request() * adding blk_fetch_request() which is combination of peek and start * disallowing completion of queued (not started) requests * applying new API to all LLDs Renamings are for consistency and to break out of tree code so that it's apparent that out of tree drivers need updating. [ Impact: block request issue API cleanup, no functional change ] Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: James Bottomley <James.Bottomley@HansenPartnership.com> Cc: Mike Miller <mike.miller@hp.com> Cc: unsik Kim <donari75@gmail.com> Cc: Paul Clements <paul.clements@steeleye.com> Cc: Tim Waugh <tim@cyberelk.net> Cc: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com> Cc: David S. Miller <davem@davemloft.net> Cc: Laurent Vivier <Laurent@lvivier.info> Cc: Jeff Garzik <jgarzik@pobox.com> Cc: Jeremy Fitzhardinge <jeremy@xensource.com> Cc: Grant Likely <grant.likely@secretlab.ca> Cc: Adrian McMenamin <adrian@mcmen.demon.co.uk> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> Cc: Borislav Petkov <petkovbb@googlemail.com> Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com> Cc: Alex Dubov <oakad@yahoo.com> Cc: Pierre Ossman <drzeus@drzeus.cx> Cc: David Woodhouse <dwmw2@infradead.org> Cc: Markus Lidel <Markus.Lidel@shadowconnect.com> Cc: Stefan Weinhuber <wein@de.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Pete Zaitcev <zaitcev@redhat.com> Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2009-05-08 02:54:16 +00:00
if (!rq)
rq = blk_fetch_request(drive->queue);
spin_unlock_irq(q->queue_lock);
spin_lock_irq(&hwif->lock);
if (!rq) {
ide_unlock_port(hwif);
goto out;
}
/*
* Sanity: don't accept a request that isn't a PM request
* if we are currently power managed. This is very important as
block: implement and enforce request peek/start/fetch Till now block layer allowed two separate modes of request execution. A request is always acquired from the request queue via elv_next_request(). After that, drivers are free to either dequeue it or process it without dequeueing. Dequeue allows elv_next_request() to return the next request so that multiple requests can be in flight. Executing requests without dequeueing has its merits mostly in allowing drivers for simpler devices which can't do sg to deal with segments only without considering request boundary. However, the benefit this brings is dubious and declining while the cost of the API ambiguity is increasing. Segment based drivers are usually for very old or limited devices and as converting to dequeueing model isn't difficult, it doesn't justify the API overhead it puts on block layer and its more modern users. Previous patches converted all block low level drivers to dequeueing model. This patch completes the API transition by... * renaming elv_next_request() to blk_peek_request() * renaming blkdev_dequeue_request() to blk_start_request() * adding blk_fetch_request() which is combination of peek and start * disallowing completion of queued (not started) requests * applying new API to all LLDs Renamings are for consistency and to break out of tree code so that it's apparent that out of tree drivers need updating. [ Impact: block request issue API cleanup, no functional change ] Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: James Bottomley <James.Bottomley@HansenPartnership.com> Cc: Mike Miller <mike.miller@hp.com> Cc: unsik Kim <donari75@gmail.com> Cc: Paul Clements <paul.clements@steeleye.com> Cc: Tim Waugh <tim@cyberelk.net> Cc: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com> Cc: David S. Miller <davem@davemloft.net> Cc: Laurent Vivier <Laurent@lvivier.info> Cc: Jeff Garzik <jgarzik@pobox.com> Cc: Jeremy Fitzhardinge <jeremy@xensource.com> Cc: Grant Likely <grant.likely@secretlab.ca> Cc: Adrian McMenamin <adrian@mcmen.demon.co.uk> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> Cc: Borislav Petkov <petkovbb@googlemail.com> Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com> Cc: Alex Dubov <oakad@yahoo.com> Cc: Pierre Ossman <drzeus@drzeus.cx> Cc: David Woodhouse <dwmw2@infradead.org> Cc: Markus Lidel <Markus.Lidel@shadowconnect.com> Cc: Stefan Weinhuber <wein@de.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Pete Zaitcev <zaitcev@redhat.com> Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2009-05-08 02:54:16 +00:00
* blk_stop_queue() doesn't prevent the blk_fetch_request()
* above to return us whatever is in the queue. Since we call
* ide_do_request() ourselves, we end up taking requests while
* the queue is blocked...
*
* We let requests forced at head of queue with ide-preempt
* though. I hope that doesn't happen too much, hopefully not
* unless the subdriver triggers such a thing in its own PM
* state machine.
*/
if ((drive->dev_flags & IDE_DFLAG_BLOCKED) &&
blk_pm_request(rq) == 0 &&
(rq->cmd_flags & REQ_PREEMPT) == 0) {
/* there should be no pending command at this point */
ide_unlock_port(hwif);
ide: don't execute the next queued command from the hard-IRQ context (v2) * Tell the block layer that we are not done handling requests by using blk_plug_device() in ide_do_request() (request handling function) and ide_timer_expiry() (timeout handler) if the queue is not empty. * Remove optimization which directly calls ide_do_request() for the next queued command from the ide_intr() (IRQ handler) and ide_timer_expiry(). * Remove no longer needed IRQ masking from ide_do_request() - in case of IDE ports needing serialization disable_irq_nosync()/enable_irq() was used for the (possibly shared) IRQ of the other IDE port. * Put the misplaced comment in the right place in ide_do_request(). * Drop no longer needed 'int masked_irq' argument from ide_do_request(). * Merge ide_do_request() into do_ide_request(). * Remove no longer needed IDE_NO_IRQ define. While at it: * Don't use HWGROUP() macro in do_ide_request(). * Use __func__ in ide_intr(). This patch reduces IRQ hadling latency for IDE and improves the system-wide handling of shared IRQs (which should result in more timeout resistant and stable IDE systems). It also makes it possible to do some further changes later (i.e. replace some busy-waiting delays with sleeping equivalents). v2: Changes per review from Elias Oltmanns: - fix wrong goto statement in 'if (startstop == ide_stopped)' block - use spin_unlock_irq() - don't use obsolete HWIF() macro Cc: Elias Oltmanns <eo@nebensachen.de> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2009-01-02 15:12:48 +00:00
goto plug_device;
}
hwif->rq = rq;
spin_unlock_irq(&hwif->lock);
startstop = start_request(drive, rq);
spin_lock_irq(&hwif->lock);
ide: don't execute the next queued command from the hard-IRQ context (v2) * Tell the block layer that we are not done handling requests by using blk_plug_device() in ide_do_request() (request handling function) and ide_timer_expiry() (timeout handler) if the queue is not empty. * Remove optimization which directly calls ide_do_request() for the next queued command from the ide_intr() (IRQ handler) and ide_timer_expiry(). * Remove no longer needed IRQ masking from ide_do_request() - in case of IDE ports needing serialization disable_irq_nosync()/enable_irq() was used for the (possibly shared) IRQ of the other IDE port. * Put the misplaced comment in the right place in ide_do_request(). * Drop no longer needed 'int masked_irq' argument from ide_do_request(). * Merge ide_do_request() into do_ide_request(). * Remove no longer needed IDE_NO_IRQ define. While at it: * Don't use HWGROUP() macro in do_ide_request(). * Use __func__ in ide_intr(). This patch reduces IRQ hadling latency for IDE and improves the system-wide handling of shared IRQs (which should result in more timeout resistant and stable IDE systems). It also makes it possible to do some further changes later (i.e. replace some busy-waiting delays with sleeping equivalents). v2: Changes per review from Elias Oltmanns: - fix wrong goto statement in 'if (startstop == ide_stopped)' block - use spin_unlock_irq() - don't use obsolete HWIF() macro Cc: Elias Oltmanns <eo@nebensachen.de> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2009-01-02 15:12:48 +00:00
if (startstop == ide_stopped) {
rq = hwif->rq;
hwif->rq = NULL;
goto repeat;
}
} else
goto plug_device;
out:
spin_unlock_irq(&hwif->lock);
if (rq == NULL)
ide_unlock_host(host);
spin_lock_irq(q->queue_lock);
ide: don't execute the next queued command from the hard-IRQ context (v2) * Tell the block layer that we are not done handling requests by using blk_plug_device() in ide_do_request() (request handling function) and ide_timer_expiry() (timeout handler) if the queue is not empty. * Remove optimization which directly calls ide_do_request() for the next queued command from the ide_intr() (IRQ handler) and ide_timer_expiry(). * Remove no longer needed IRQ masking from ide_do_request() - in case of IDE ports needing serialization disable_irq_nosync()/enable_irq() was used for the (possibly shared) IRQ of the other IDE port. * Put the misplaced comment in the right place in ide_do_request(). * Drop no longer needed 'int masked_irq' argument from ide_do_request(). * Merge ide_do_request() into do_ide_request(). * Remove no longer needed IDE_NO_IRQ define. While at it: * Don't use HWGROUP() macro in do_ide_request(). * Use __func__ in ide_intr(). This patch reduces IRQ hadling latency for IDE and improves the system-wide handling of shared IRQs (which should result in more timeout resistant and stable IDE systems). It also makes it possible to do some further changes later (i.e. replace some busy-waiting delays with sleeping equivalents). v2: Changes per review from Elias Oltmanns: - fix wrong goto statement in 'if (startstop == ide_stopped)' block - use spin_unlock_irq() - don't use obsolete HWIF() macro Cc: Elias Oltmanns <eo@nebensachen.de> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2009-01-02 15:12:48 +00:00
return;
ide: don't execute the next queued command from the hard-IRQ context (v2) * Tell the block layer that we are not done handling requests by using blk_plug_device() in ide_do_request() (request handling function) and ide_timer_expiry() (timeout handler) if the queue is not empty. * Remove optimization which directly calls ide_do_request() for the next queued command from the ide_intr() (IRQ handler) and ide_timer_expiry(). * Remove no longer needed IRQ masking from ide_do_request() - in case of IDE ports needing serialization disable_irq_nosync()/enable_irq() was used for the (possibly shared) IRQ of the other IDE port. * Put the misplaced comment in the right place in ide_do_request(). * Drop no longer needed 'int masked_irq' argument from ide_do_request(). * Merge ide_do_request() into do_ide_request(). * Remove no longer needed IDE_NO_IRQ define. While at it: * Don't use HWGROUP() macro in do_ide_request(). * Use __func__ in ide_intr(). This patch reduces IRQ hadling latency for IDE and improves the system-wide handling of shared IRQs (which should result in more timeout resistant and stable IDE systems). It also makes it possible to do some further changes later (i.e. replace some busy-waiting delays with sleeping equivalents). v2: Changes per review from Elias Oltmanns: - fix wrong goto statement in 'if (startstop == ide_stopped)' block - use spin_unlock_irq() - don't use obsolete HWIF() macro Cc: Elias Oltmanns <eo@nebensachen.de> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2009-01-02 15:12:48 +00:00
plug_device:
spin_unlock_irq(&hwif->lock);
ide_unlock_host(host);
plug_device_2:
spin_lock_irq(q->queue_lock);
if (rq)
blk_requeue_request(q, rq);
if (!elv_queue_empty(q))
blk_plug_device(q);
}
static void ide_requeue_and_plug(ide_drive_t *drive, struct request *rq)
{
struct request_queue *q = drive->queue;
unsigned long flags;
spin_lock_irqsave(q->queue_lock, flags);
if (rq)
blk_requeue_request(q, rq);
if (!elv_queue_empty(q))
blk_plug_device(q);
spin_unlock_irqrestore(q->queue_lock, flags);
}
static int drive_is_ready(ide_drive_t *drive)
{
ide_hwif_t *hwif = drive->hwif;
u8 stat = 0;
if (drive->waiting_for_dma)
return hwif->dma_ops->dma_test_irq(drive);
if (hwif->io_ports.ctl_addr &&
(hwif->host_flags & IDE_HFLAG_BROKEN_ALTSTATUS) == 0)
stat = hwif->tp_ops->read_altstatus(hwif);
else
/* Note: this may clear a pending IRQ!! */
stat = hwif->tp_ops->read_status(hwif);
if (stat & ATA_BUSY)
/* drive busy: definitely not interrupting */
return 0;
/* drive ready: *might* be interrupting */
return 1;
}
/**
* ide_timer_expiry - handle lack of an IDE interrupt
* @data: timer callback magic (hwif)
*
* An IDE command has timed out before the expected drive return
* occurred. At this point we attempt to clean up the current
* mess. If the current handler includes an expiry handler then
* we invoke the expiry handler, and providing it is happy the
* work is done. If that fails we apply generic recovery rules
* invoking the handler and checking the drive DMA status. We
* have an excessively incestuous relationship with the DMA
* logic that wants cleaning up.
*/
void ide_timer_expiry (unsigned long data)
{
ide_hwif_t *hwif = (ide_hwif_t *)data;
ide_drive_t *uninitialized_var(drive);
ide_handler_t *handler;
unsigned long flags;
int wait = -1;
int plug_device = 0;
struct request *uninitialized_var(rq_in_flight);
spin_lock_irqsave(&hwif->lock, flags);
handler = hwif->handler;
if (handler == NULL || hwif->req_gen != hwif->req_gen_timer) {
/*
* Either a marginal timeout occurred
* (got the interrupt just as timer expired),
* or we were "sleeping" to give other devices a chance.
* Either way, we don't really want to complain about anything.
*/
} else {
ide_expiry_t *expiry = hwif->expiry;
ide_startstop_t startstop = ide_stopped;
drive = hwif->cur_dev;
if (expiry) {
wait = expiry(drive);
if (wait > 0) { /* continue */
/* reset timer */
hwif->timer.expires = jiffies + wait;
hwif->req_gen_timer = hwif->req_gen;
add_timer(&hwif->timer);
spin_unlock_irqrestore(&hwif->lock, flags);
return;
ide: don't execute the next queued command from the hard-IRQ context (v2) * Tell the block layer that we are not done handling requests by using blk_plug_device() in ide_do_request() (request handling function) and ide_timer_expiry() (timeout handler) if the queue is not empty. * Remove optimization which directly calls ide_do_request() for the next queued command from the ide_intr() (IRQ handler) and ide_timer_expiry(). * Remove no longer needed IRQ masking from ide_do_request() - in case of IDE ports needing serialization disable_irq_nosync()/enable_irq() was used for the (possibly shared) IRQ of the other IDE port. * Put the misplaced comment in the right place in ide_do_request(). * Drop no longer needed 'int masked_irq' argument from ide_do_request(). * Merge ide_do_request() into do_ide_request(). * Remove no longer needed IDE_NO_IRQ define. While at it: * Don't use HWGROUP() macro in do_ide_request(). * Use __func__ in ide_intr(). This patch reduces IRQ hadling latency for IDE and improves the system-wide handling of shared IRQs (which should result in more timeout resistant and stable IDE systems). It also makes it possible to do some further changes later (i.e. replace some busy-waiting delays with sleeping equivalents). v2: Changes per review from Elias Oltmanns: - fix wrong goto statement in 'if (startstop == ide_stopped)' block - use spin_unlock_irq() - don't use obsolete HWIF() macro Cc: Elias Oltmanns <eo@nebensachen.de> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
2009-01-02 15:12:48 +00:00
}
}
hwif->handler = NULL;
hwif->expiry = NULL;
/*
* We need to simulate a real interrupt when invoking
* the handler() function, which means we need to
* globally mask the specific IRQ:
*/
spin_unlock(&hwif->lock);
/* disable_irq_nosync ?? */
disable_irq(hwif->irq);
/* local CPU only, as if we were handling an interrupt */
local_irq_disable();
if (hwif->polling) {
startstop = handler(drive);
} else if (drive_is_ready(drive)) {
if (drive->waiting_for_dma)
hwif->dma_ops->dma_lost_irq(drive);
if (hwif->port_ops && hwif->port_ops->clear_irq)
hwif->port_ops->clear_irq(drive);
printk(KERN_WARNING "%s: lost interrupt\n",
drive->name);
startstop = handler(drive);
} else {
if (drive->waiting_for_dma)
startstop = ide_dma_timeout_retry(drive, wait);
else
startstop = ide_error(drive, "irq timeout",
hwif->tp_ops->read_status(hwif));
}
spin_lock_irq(&hwif->lock);
enable_irq(hwif->irq);
if (startstop == ide_stopped && hwif->polling == 0) {
rq_in_flight = hwif->rq;
hwif->rq = NULL;
ide_unlock_port(hwif);
plug_device = 1;
}
}
spin_unlock_irqrestore(&hwif->lock, flags);
if (plug_device) {
ide_unlock_host(hwif->host);
ide_requeue_and_plug(drive, rq_in_flight);
}
}
/**
* unexpected_intr - handle an unexpected IDE interrupt
* @irq: interrupt line
* @hwif: port being processed
*
* There's nothing really useful we can do with an unexpected interrupt,
* other than reading the status register (to clear it), and logging it.
* There should be no way that an irq can happen before we're ready for it,
* so we needn't worry much about losing an "important" interrupt here.
*
* On laptops (and "green" PCs), an unexpected interrupt occurs whenever
* the drive enters "idle", "standby", or "sleep" mode, so if the status
* looks "good", we just ignore the interrupt completely.
*
* This routine assumes __cli() is in effect when called.
*
* If an unexpected interrupt happens on irq15 while we are handling irq14
* and if the two interfaces are "serialized" (CMD640), then it looks like
* we could screw up by interfering with a new request being set up for
* irq15.
*
* In reality, this is a non-issue. The new command is not sent unless
* the drive is ready to accept one, in which case we know the drive is
* not trying to interrupt us. And ide_set_handler() is always invoked
* before completing the issuance of any new drive command, so we will not
* be accidentally invoked as a result of any valid command completion
* interrupt.
*/
static void unexpected_intr(int irq, ide_hwif_t *hwif)
{
u8 stat = hwif->tp_ops->read_status(hwif);
if (!OK_STAT(stat, ATA_DRDY, BAD_STAT)) {
/* Try to not flood the console with msgs */
static unsigned long last_msgtime, count;
++count;
if (time_after(jiffies, last_msgtime + HZ)) {
last_msgtime = jiffies;
printk(KERN_ERR "%s: unexpected interrupt, "
"status=0x%02x, count=%ld\n",
hwif->name, stat, count);
}
}
}
/**
* ide_intr - default IDE interrupt handler
* @irq: interrupt number
* @dev_id: hwif
* @regs: unused weirdness from the kernel irq layer
*
* This is the default IRQ handler for the IDE layer. You should
* not need to override it. If you do be aware it is subtle in
* places
*
* hwif is the interface in the group currently performing
* a command. hwif->cur_dev is the drive and hwif->handler is
* the IRQ handler to call. As we issue a command the handlers
* step through multiple states, reassigning the handler to the
* next step in the process. Unlike a smart SCSI controller IDE
* expects the main processor to sequence the various transfer
* stages. We also manage a poll timer to catch up with most
* timeout situations. There are still a few where the handlers
* don't ever decide to give up.
*
* The handler eventually returns ide_stopped to indicate the
* request completed. At this point we issue the next request
* on the port and the process begins again.
*/
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
irqreturn_t ide_intr (int irq, void *dev_id)
{
ide_hwif_t *hwif = (ide_hwif_t *)dev_id;
struct ide_host *host = hwif->host;
ide_drive_t *uninitialized_var(drive);
ide_handler_t *handler;
unsigned long flags;
ide_startstop_t startstop;
irqreturn_t irq_ret = IRQ_NONE;
int plug_device = 0;
struct request *uninitialized_var(rq_in_flight);
if (host->host_flags & IDE_HFLAG_SERIALIZE) {
if (hwif != host->cur_port)
goto out_early;
}
spin_lock_irqsave(&hwif->lock, flags);
if (hwif->ack_intr && hwif->ack_intr(hwif) == 0)
goto out;
handler = hwif->handler;
if (handler == NULL || hwif->polling) {
/*
* Not expecting an interrupt from this drive.
* That means this could be:
* (1) an interrupt from another PCI device
* sharing the same PCI INT# as us.
* or (2) a drive just entered sleep or standby mode,
* and is interrupting to let us know.
* or (3) a spurious interrupt of unknown origin.
*
* For PCI, we cannot tell the difference,
* so in that case we just ignore it and hope it goes away.
*/
if ((host->irq_flags & IRQF_SHARED) == 0) {
/*
* Probably not a shared PCI interrupt,
* so we can safely try to do something about it:
*/
unexpected_intr(irq, hwif);
} else {
/*
* Whack the status register, just in case
* we have a leftover pending IRQ.
*/
(void)hwif->tp_ops->read_status(hwif);
}
goto out;
}
drive = hwif->cur_dev;
if (!drive_is_ready(drive))
/*
* This happens regularly when we share a PCI IRQ with
* another device. Unfortunately, it can also happen
* with some buggy drives that trigger the IRQ before
* their status register is up to date. Hopefully we have
* enough advance overhead that the latter isn't a problem.
*/
goto out;
hwif->handler = NULL;
hwif->expiry = NULL;
hwif->req_gen++;
del_timer(&hwif->timer);
spin_unlock(&hwif->lock);
if (hwif->port_ops && hwif->port_ops->clear_irq)
hwif->port_ops->clear_irq(drive);
if (drive->dev_flags & IDE_DFLAG_UNMASK)
local_irq_enable_in_hardirq();
/* service this interrupt, may set handler for next interrupt */
startstop = handler(drive);
spin_lock_irq(&hwif->lock);
/*
* Note that handler() may have set things up for another
* interrupt to occur soon, but it cannot happen until
* we exit from this routine, because it will be the
* same irq as is currently being serviced here, and Linux
* won't allow another of the same (on any CPU) until we return.
*/
if (startstop == ide_stopped && hwif->polling == 0) {
BUG_ON(hwif->handler);
rq_in_flight = hwif->rq;
hwif->rq = NULL;
ide_unlock_port(hwif);
plug_device = 1;
}
irq_ret = IRQ_HANDLED;
out:
spin_unlock_irqrestore(&hwif->lock, flags);
out_early:
if (plug_device) {
ide_unlock_host(hwif->host);
ide_requeue_and_plug(drive, rq_in_flight);
}
return irq_ret;
}
EXPORT_SYMBOL_GPL(ide_intr);
void ide_pad_transfer(ide_drive_t *drive, int write, int len)
{
ide_hwif_t *hwif = drive->hwif;
u8 buf[4] = { 0 };
while (len > 0) {
if (write)
hwif->tp_ops->output_data(drive, NULL, buf, min(4, len));
else
hwif->tp_ops->input_data(drive, NULL, buf, min(4, len));
len -= 4;
}
}
EXPORT_SYMBOL_GPL(ide_pad_transfer);