linux/drivers/ieee1394/sbp2.c

2204 lines
68 KiB
C
Raw Normal View History

/*
* sbp2.c - SBP-2 protocol driver for IEEE-1394
*
* Copyright (C) 2000 James Goodwin, Filanet Corporation (www.filanet.com)
* jamesg@filanet.com (JSG)
*
* Copyright (C) 2003 Ben Collins <bcollins@debian.org>
*
* 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 of the License, 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.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*/
/*
* Brief Description:
*
* This driver implements the Serial Bus Protocol 2 (SBP-2) over IEEE-1394
* under Linux. The SBP-2 driver is implemented as an IEEE-1394 high-level
* driver. It also registers as a SCSI lower-level driver in order to accept
* SCSI commands for transport using SBP-2.
*
* You may access any attached SBP-2 storage devices as if they were SCSI
* devices (e.g. mount /dev/sda1, fdisk, mkfs, etc.).
*
* Current Issues:
*
* - Error Handling: SCSI aborts and bus reset requests are handled somewhat
* but the code needs additional debugging.
*/
#include <linux/blkdev.h>
#include <linux/compiler.h>
#include <linux/delay.h>
#include <linux/device.h>
#include <linux/dma-mapping.h>
#include <linux/gfp.h>
#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <linux/pci.h>
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/stat.h>
#include <linux/string.h>
#include <linux/stringify.h>
#include <linux/types.h>
#include <linux/wait.h>
#include <asm/byteorder.h>
#include <asm/errno.h>
#include <asm/param.h>
#include <asm/scatterlist.h>
#include <asm/system.h>
#include <asm/types.h>
#ifdef CONFIG_IEEE1394_SBP2_PHYS_DMA
#include <asm/io.h> /* for bus_to_virt */
#endif
#include <scsi/scsi.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_dbg.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_host.h>
#include "csr1212.h"
#include "highlevel.h"
#include "hosts.h"
#include "ieee1394.h"
#include "ieee1394_core.h"
#include "ieee1394_hotplug.h"
#include "ieee1394_transactions.h"
#include "ieee1394_types.h"
#include "nodemgr.h"
#include "sbp2.h"
/*
* Module load parameter definitions
*/
/*
* Change max_speed on module load if you have a bad IEEE-1394
* controller that has trouble running 2KB packets at 400mb.
*
* NOTE: On certain OHCI parts I have seen short packets on async transmit
* (probably due to PCI latency/throughput issues with the part). You can
* bump down the speed if you are running into problems.
*/
static int max_speed = IEEE1394_SPEED_MAX;
module_param(max_speed, int, 0644);
MODULE_PARM_DESC(max_speed, "Force max speed (3 = 800mb, 2 = 400mb, 1 = 200mb, 0 = 100mb)");
/*
* Set serialize_io to 1 if you'd like only one scsi command sent
* down to us at a time (debugging). This might be necessary for very
* badly behaved sbp2 devices.
*
* TODO: Make this configurable per device.
*/
static int serialize_io = 1;
module_param(serialize_io, int, 0444);
MODULE_PARM_DESC(serialize_io, "Serialize I/O coming from scsi drivers (default = 1, faster = 0)");
/*
* Bump up max_sectors if you'd like to support very large sized
* transfers. Please note that some older sbp2 bridge chips are broken for
* transfers greater or equal to 128KB. Default is a value of 255
* sectors, or just under 128KB (at 512 byte sector size). I can note that
* the Oxsemi sbp2 chipsets have no problems supporting very large
* transfer sizes.
*/
static int max_sectors = SBP2_MAX_SECTORS;
module_param(max_sectors, int, 0444);
MODULE_PARM_DESC(max_sectors, "Change max sectors per I/O supported (default = "
__stringify(SBP2_MAX_SECTORS) ")");
/*
* Exclusive login to sbp2 device? In most cases, the sbp2 driver should
* do an exclusive login, as it's generally unsafe to have two hosts
* talking to a single sbp2 device at the same time (filesystem coherency,
* etc.). If you're running an sbp2 device that supports multiple logins,
* and you're either running read-only filesystems or some sort of special
* filesystem supporting multiple hosts, e.g. OpenGFS, Oracle Cluster
* File System, or Lustre, then set exclusive_login to zero.
*
* So far only bridges from Oxford Semiconductor are known to support
* concurrent logins. Depending on firmware, four or two concurrent logins
* are possible on OXFW911 and newer Oxsemi bridges.
*/
static int exclusive_login = 1;
module_param(exclusive_login, int, 0644);
MODULE_PARM_DESC(exclusive_login, "Exclusive login to sbp2 device (default = 1)");
/*
* If any of the following workarounds is required for your device to work,
* please submit the kernel messages logged by sbp2 to the linux1394-devel
* mailing list.
*
* - 128kB max transfer
* Limit transfer size. Necessary for some old bridges.
*
* - 36 byte inquiry
* When scsi_mod probes the device, let the inquiry command look like that
* from MS Windows.
*
* - skip mode page 8
* Suppress sending of mode_sense for mode page 8 if the device pretends to
* support the SCSI Primary Block commands instead of Reduced Block Commands.
*
* - fix capacity
* Tell sd_mod to correct the last sector number reported by read_capacity.
* Avoids access beyond actual disk limits on devices with an off-by-one bug.
* Don't use this with devices which don't have this bug.
*
* - override internal blacklist
* Instead of adding to the built-in blacklist, use only the workarounds
* specified in the module load parameter.
* Useful if a blacklist entry interfered with a non-broken device.
*/
static int sbp2_default_workarounds;
module_param_named(workarounds, sbp2_default_workarounds, int, 0644);
MODULE_PARM_DESC(workarounds, "Work around device bugs (default = 0"
", 128kB max transfer = " __stringify(SBP2_WORKAROUND_128K_MAX_TRANS)
", 36 byte inquiry = " __stringify(SBP2_WORKAROUND_INQUIRY_36)
", skip mode page 8 = " __stringify(SBP2_WORKAROUND_MODE_SENSE_8)
", fix capacity = " __stringify(SBP2_WORKAROUND_FIX_CAPACITY)
", override internal blacklist = " __stringify(SBP2_WORKAROUND_OVERRIDE)
", or a combination)");
#define SBP2_INFO(fmt, args...) HPSB_INFO("sbp2: "fmt, ## args)
#define SBP2_ERR(fmt, args...) HPSB_ERR("sbp2: "fmt, ## args)
/*
* Globals
*/
static void sbp2scsi_complete_all_commands(struct scsi_id_instance_data *, u32);
static void sbp2scsi_complete_command(struct scsi_id_instance_data *, u32,
struct scsi_cmnd *,
void (*)(struct scsi_cmnd *));
static struct scsi_id_instance_data *sbp2_alloc_device(struct unit_directory *);
static int sbp2_start_device(struct scsi_id_instance_data *);
static void sbp2_remove_device(struct scsi_id_instance_data *);
static int sbp2_login_device(struct scsi_id_instance_data *);
static int sbp2_reconnect_device(struct scsi_id_instance_data *);
static int sbp2_logout_device(struct scsi_id_instance_data *);
static void sbp2_host_reset(struct hpsb_host *);
static int sbp2_handle_status_write(struct hpsb_host *, int, int, quadlet_t *,
u64, size_t, u16);
static int sbp2_agent_reset(struct scsi_id_instance_data *, int);
static void sbp2_parse_unit_directory(struct scsi_id_instance_data *,
struct unit_directory *);
static int sbp2_set_busy_timeout(struct scsi_id_instance_data *);
static int sbp2_max_speed_and_size(struct scsi_id_instance_data *);
static const u8 sbp2_speedto_max_payload[] = { 0x7, 0x8, 0x9, 0xA, 0xB, 0xC };
static struct hpsb_highlevel sbp2_highlevel = {
.name = SBP2_DEVICE_NAME,
.host_reset = sbp2_host_reset,
};
static struct hpsb_address_ops sbp2_ops = {
.write = sbp2_handle_status_write
};
#ifdef CONFIG_IEEE1394_SBP2_PHYS_DMA
static int sbp2_handle_physdma_write(struct hpsb_host *, int, int, quadlet_t *,
u64, size_t, u16);
static int sbp2_handle_physdma_read(struct hpsb_host *, int, quadlet_t *, u64,
size_t, u16);
static struct hpsb_address_ops sbp2_physdma_ops = {
.read = sbp2_handle_physdma_read,
.write = sbp2_handle_physdma_write,
};
#endif
/*
* Interface to driver core and IEEE 1394 core
*/
static struct ieee1394_device_id sbp2_id_table[] = {
{
.match_flags = IEEE1394_MATCH_SPECIFIER_ID | IEEE1394_MATCH_VERSION,
.specifier_id = SBP2_UNIT_SPEC_ID_ENTRY & 0xffffff,
.version = SBP2_SW_VERSION_ENTRY & 0xffffff},
{}
};
MODULE_DEVICE_TABLE(ieee1394, sbp2_id_table);
static int sbp2_probe(struct device *);
static int sbp2_remove(struct device *);
static int sbp2_update(struct unit_directory *);
static struct hpsb_protocol_driver sbp2_driver = {
.name = "SBP2 Driver",
.id_table = sbp2_id_table,
.update = sbp2_update,
.driver = {
.name = SBP2_DEVICE_NAME,
.bus = &ieee1394_bus_type,
.probe = sbp2_probe,
.remove = sbp2_remove,
},
};
/*
* Interface to SCSI core
*/
static int sbp2scsi_queuecommand(struct scsi_cmnd *,
void (*)(struct scsi_cmnd *));
static int sbp2scsi_abort(struct scsi_cmnd *);
static int sbp2scsi_reset(struct scsi_cmnd *);
static int sbp2scsi_slave_alloc(struct scsi_device *);
static int sbp2scsi_slave_configure(struct scsi_device *);
static void sbp2scsi_slave_destroy(struct scsi_device *);
static ssize_t sbp2_sysfs_ieee1394_id_show(struct device *,
struct device_attribute *, char *);
static DEVICE_ATTR(ieee1394_id, S_IRUGO, sbp2_sysfs_ieee1394_id_show, NULL);
static struct device_attribute *sbp2_sysfs_sdev_attrs[] = {
&dev_attr_ieee1394_id,
NULL
};
static struct scsi_host_template scsi_driver_template = {
.module = THIS_MODULE,
.name = "SBP-2 IEEE-1394",
.proc_name = SBP2_DEVICE_NAME,
.queuecommand = sbp2scsi_queuecommand,
.eh_abort_handler = sbp2scsi_abort,
.eh_device_reset_handler = sbp2scsi_reset,
.slave_alloc = sbp2scsi_slave_alloc,
.slave_configure = sbp2scsi_slave_configure,
.slave_destroy = sbp2scsi_slave_destroy,
.this_id = -1,
.sg_tablesize = SG_ALL,
.use_clustering = ENABLE_CLUSTERING,
.cmd_per_lun = SBP2_MAX_CMDS,
.can_queue = SBP2_MAX_CMDS,
.emulated = 1,
.sdev_attrs = sbp2_sysfs_sdev_attrs,
};
/*
* List of devices with known bugs.
*
* The firmware_revision field, masked with 0xffff00, is the best indicator
* for the type of bridge chip of a device. It yields a few false positives
* but this did not break correctly behaving devices so far.
*/
static const struct {
u32 firmware_revision;
u32 model_id;
unsigned workarounds;
} sbp2_workarounds_table[] = {
/* DViCO Momobay CX-1 with TSB42AA9 bridge */ {
.firmware_revision = 0x002800,
.model_id = 0x001010,
.workarounds = SBP2_WORKAROUND_INQUIRY_36 |
SBP2_WORKAROUND_MODE_SENSE_8,
},
/* Initio bridges, actually only needed for some older ones */ {
.firmware_revision = 0x000200,
.workarounds = SBP2_WORKAROUND_INQUIRY_36,
},
/* Symbios bridge */ {
.firmware_revision = 0xa0b800,
.workarounds = SBP2_WORKAROUND_128K_MAX_TRANS,
},
/*
* Note about the following Apple iPod blacklist entries:
*
* There are iPods (2nd gen, 3rd gen) with model_id==0. Since our
* matching logic treats 0 as a wildcard, we cannot match this ID
* without rewriting the matching routine. Fortunately these iPods
* do not feature the read_capacity bug according to one report.
* Read_capacity behaviour as well as model_id could change due to
* Apple-supplied firmware updates though.
*/
/* iPod 4th generation */ {
.firmware_revision = 0x0a2700,
.model_id = 0x000021,
.workarounds = SBP2_WORKAROUND_FIX_CAPACITY,
},
/* iPod mini */ {
.firmware_revision = 0x0a2700,
.model_id = 0x000023,
.workarounds = SBP2_WORKAROUND_FIX_CAPACITY,
},
/* iPod Photo */ {
.firmware_revision = 0x0a2700,
.model_id = 0x00007e,
.workarounds = SBP2_WORKAROUND_FIX_CAPACITY,
}
};
/**************************************
* General utility functions
**************************************/
#ifndef __BIG_ENDIAN
/*
* Converts a buffer from be32 to cpu byte ordering. Length is in bytes.
*/
static inline void sbp2util_be32_to_cpu_buffer(void *buffer, int length)
{
u32 *temp = buffer;
for (length = (length >> 2); length--; )
temp[length] = be32_to_cpu(temp[length]);
return;
}
/*
* Converts a buffer from cpu to be32 byte ordering. Length is in bytes.
*/
static inline void sbp2util_cpu_to_be32_buffer(void *buffer, int length)
{
u32 *temp = buffer;
for (length = (length >> 2); length--; )
temp[length] = cpu_to_be32(temp[length]);
return;
}
#else /* BIG_ENDIAN */
/* Why waste the cpu cycles? */
#define sbp2util_be32_to_cpu_buffer(x,y) do {} while (0)
#define sbp2util_cpu_to_be32_buffer(x,y) do {} while (0)
#endif
static DECLARE_WAIT_QUEUE_HEAD(access_wq);
/*
* Waits for completion of an SBP-2 access request.
* Returns nonzero if timed out or prematurely interrupted.
*/
static int sbp2util_access_timeout(struct scsi_id_instance_data *scsi_id,
int timeout)
{
long leftover = wait_event_interruptible_timeout(
access_wq, scsi_id->access_complete, timeout);
scsi_id->access_complete = 0;
return leftover <= 0;
}
static void sbp2_free_packet(struct hpsb_packet *packet)
{
hpsb_free_tlabel(packet);
hpsb_free_packet(packet);
}
/*
* This is much like hpsb_node_write(), except it ignores the response
* subaction and returns immediately. Can be used from atomic context.
*/
static int sbp2util_node_write_no_wait(struct node_entry *ne, u64 addr,
quadlet_t *buffer, size_t length)
{
struct hpsb_packet *packet;
packet = hpsb_make_writepacket(ne->host, ne->nodeid,
addr, buffer, length);
if (!packet)
return -ENOMEM;
hpsb_set_packet_complete_task(packet,
(void (*)(void *))sbp2_free_packet,
packet);
hpsb_node_fill_packet(ne, packet);
if (hpsb_send_packet(packet) < 0) {
sbp2_free_packet(packet);
return -EIO;
}
return 0;
}
static void sbp2util_notify_fetch_agent(struct scsi_id_instance_data *scsi_id,
u64 offset, quadlet_t *data, size_t len)
{
/*
* There is a small window after a bus reset within which the node
* entry's generation is current but the reconnect wasn't completed.
*/
if (unlikely(atomic_read(&scsi_id->state) == SBP2LU_STATE_IN_RESET))
return;
if (hpsb_node_write(scsi_id->ne,
scsi_id->sbp2_command_block_agent_addr + offset,
data, len))
SBP2_ERR("sbp2util_notify_fetch_agent failed.");
/*
* Now accept new SCSI commands, unless a bus reset happended during
* hpsb_node_write.
*/
if (likely(atomic_read(&scsi_id->state) != SBP2LU_STATE_IN_RESET))
scsi_unblock_requests(scsi_id->scsi_host);
}
static void sbp2util_write_orb_pointer(struct work_struct *work)
{
quadlet_t data[2];
data[0] = ORB_SET_NODE_ID(
(container_of(work, struct scsi_id_instance_data, protocol_work))->hi->host->node_id);
data[1] = (container_of(work, struct scsi_id_instance_data, protocol_work))->last_orb_dma;
sbp2util_cpu_to_be32_buffer(data, 8);
sbp2util_notify_fetch_agent(container_of(work, struct scsi_id_instance_data, protocol_work), SBP2_ORB_POINTER_OFFSET, data, 8);
}
static void sbp2util_write_doorbell(struct work_struct *work)
{
sbp2util_notify_fetch_agent(container_of(work, struct scsi_id_instance_data, protocol_work), SBP2_DOORBELL_OFFSET, NULL, 4);
}
static int sbp2util_create_command_orb_pool(struct scsi_id_instance_data *scsi_id)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
int i;
unsigned long flags, orbs;
struct sbp2_command_info *command;
orbs = serialize_io ? 2 : SBP2_MAX_CMDS;
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
for (i = 0; i < orbs; i++) {
command = kzalloc(sizeof(*command), GFP_ATOMIC);
if (!command) {
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock,
flags);
return -ENOMEM;
}
command->command_orb_dma =
pci_map_single(hi->host->pdev, &command->command_orb,
sizeof(struct sbp2_command_orb),
PCI_DMA_TODEVICE);
command->sge_dma =
pci_map_single(hi->host->pdev,
&command->scatter_gather_element,
sizeof(command->scatter_gather_element),
PCI_DMA_BIDIRECTIONAL);
INIT_LIST_HEAD(&command->list);
list_add_tail(&command->list, &scsi_id->sbp2_command_orb_completed);
}
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
return 0;
}
static void sbp2util_remove_command_orb_pool(struct scsi_id_instance_data *scsi_id)
{
struct hpsb_host *host = scsi_id->hi->host;
struct list_head *lh, *next;
struct sbp2_command_info *command;
unsigned long flags;
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
if (!list_empty(&scsi_id->sbp2_command_orb_completed)) {
list_for_each_safe(lh, next, &scsi_id->sbp2_command_orb_completed) {
command = list_entry(lh, struct sbp2_command_info, list);
pci_unmap_single(host->pdev, command->command_orb_dma,
sizeof(struct sbp2_command_orb),
PCI_DMA_TODEVICE);
pci_unmap_single(host->pdev, command->sge_dma,
sizeof(command->scatter_gather_element),
PCI_DMA_BIDIRECTIONAL);
kfree(command);
}
}
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
return;
}
/*
* Finds the sbp2_command for a given outstanding command ORB.
* Only looks at the in-use list.
*/
static struct sbp2_command_info *sbp2util_find_command_for_orb(
struct scsi_id_instance_data *scsi_id, dma_addr_t orb)
{
struct sbp2_command_info *command;
unsigned long flags;
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
if (!list_empty(&scsi_id->sbp2_command_orb_inuse)) {
list_for_each_entry(command, &scsi_id->sbp2_command_orb_inuse, list) {
if (command->command_orb_dma == orb) {
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
return command;
}
}
}
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
return NULL;
}
/*
* Finds the sbp2_command for a given outstanding SCpnt.
* Only looks at the in-use list.
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
* Must be called with scsi_id->sbp2_command_orb_lock held.
*/
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
static struct sbp2_command_info *sbp2util_find_command_for_SCpnt(
struct scsi_id_instance_data *scsi_id, void *SCpnt)
{
struct sbp2_command_info *command;
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
if (!list_empty(&scsi_id->sbp2_command_orb_inuse))
list_for_each_entry(command, &scsi_id->sbp2_command_orb_inuse, list)
if (command->Current_SCpnt == SCpnt)
return command;
return NULL;
}
static struct sbp2_command_info *sbp2util_allocate_command_orb(
struct scsi_id_instance_data *scsi_id,
struct scsi_cmnd *Current_SCpnt,
void (*Current_done)(struct scsi_cmnd *))
{
struct list_head *lh;
struct sbp2_command_info *command = NULL;
unsigned long flags;
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
if (!list_empty(&scsi_id->sbp2_command_orb_completed)) {
lh = scsi_id->sbp2_command_orb_completed.next;
list_del(lh);
command = list_entry(lh, struct sbp2_command_info, list);
command->Current_done = Current_done;
command->Current_SCpnt = Current_SCpnt;
list_add_tail(&command->list, &scsi_id->sbp2_command_orb_inuse);
} else {
SBP2_ERR("%s: no orbs available", __FUNCTION__);
}
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
return command;
}
static void sbp2util_free_command_dma(struct sbp2_command_info *command)
{
struct scsi_id_instance_data *scsi_id =
(struct scsi_id_instance_data *)command->Current_SCpnt->device->host->hostdata[0];
struct hpsb_host *host;
if (!scsi_id) {
SBP2_ERR("%s: scsi_id == NULL", __FUNCTION__);
return;
}
host = scsi_id->ud->ne->host;
if (command->cmd_dma) {
if (command->dma_type == CMD_DMA_SINGLE)
pci_unmap_single(host->pdev, command->cmd_dma,
command->dma_size, command->dma_dir);
else if (command->dma_type == CMD_DMA_PAGE)
pci_unmap_page(host->pdev, command->cmd_dma,
command->dma_size, command->dma_dir);
/* XXX: Check for CMD_DMA_NONE bug */
command->dma_type = CMD_DMA_NONE;
command->cmd_dma = 0;
}
if (command->sge_buffer) {
pci_unmap_sg(host->pdev, command->sge_buffer,
command->dma_size, command->dma_dir);
command->sge_buffer = NULL;
}
}
/*
* This function moves a command to the completed orb list.
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
* Must be called with scsi_id->sbp2_command_orb_lock held.
*/
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
static void sbp2util_mark_command_completed(
struct scsi_id_instance_data *scsi_id,
struct sbp2_command_info *command)
{
list_del(&command->list);
sbp2util_free_command_dma(command);
list_add_tail(&command->list, &scsi_id->sbp2_command_orb_completed);
}
/*
* Is scsi_id valid? Is the 1394 node still present?
*/
static inline int sbp2util_node_is_available(struct scsi_id_instance_data *scsi_id)
{
return scsi_id && scsi_id->ne && !scsi_id->ne->in_limbo;
}
/*********************************************
* IEEE-1394 core driver stack related section
*********************************************/
static int sbp2_probe(struct device *dev)
{
struct unit_directory *ud;
struct scsi_id_instance_data *scsi_id;
ud = container_of(dev, struct unit_directory, device);
/* Don't probe UD's that have the LUN flag. We'll probe the LUN(s)
* instead. */
if (ud->flags & UNIT_DIRECTORY_HAS_LUN_DIRECTORY)
return -ENODEV;
scsi_id = sbp2_alloc_device(ud);
if (!scsi_id)
return -ENOMEM;
sbp2_parse_unit_directory(scsi_id, ud);
return sbp2_start_device(scsi_id);
}
static int sbp2_remove(struct device *dev)
{
struct unit_directory *ud;
struct scsi_id_instance_data *scsi_id;
struct scsi_device *sdev;
ud = container_of(dev, struct unit_directory, device);
scsi_id = ud->device.driver_data;
if (!scsi_id)
return 0;
if (scsi_id->scsi_host) {
/* Get rid of enqueued commands if there is no chance to
* send them. */
if (!sbp2util_node_is_available(scsi_id))
sbp2scsi_complete_all_commands(scsi_id, DID_NO_CONNECT);
/* scsi_remove_device() may trigger shutdown functions of SCSI
* highlevel drivers which would deadlock if blocked. */
atomic_set(&scsi_id->state, SBP2LU_STATE_IN_SHUTDOWN);
scsi_unblock_requests(scsi_id->scsi_host);
}
sdev = scsi_id->sdev;
if (sdev) {
scsi_id->sdev = NULL;
scsi_remove_device(sdev);
}
sbp2_logout_device(scsi_id);
sbp2_remove_device(scsi_id);
return 0;
}
static int sbp2_update(struct unit_directory *ud)
{
struct scsi_id_instance_data *scsi_id = ud->device.driver_data;
if (sbp2_reconnect_device(scsi_id)) {
/* Reconnect has failed. Perhaps we didn't reconnect fast
* enough. Try a regular login, but first log out just in
* case of any weirdness. */
sbp2_logout_device(scsi_id);
if (sbp2_login_device(scsi_id)) {
/* Login failed too, just fail, and the backend
* will call our sbp2_remove for us */
SBP2_ERR("Failed to reconnect to sbp2 device!");
return -EBUSY;
}
}
sbp2_set_busy_timeout(scsi_id);
sbp2_agent_reset(scsi_id, 1);
sbp2_max_speed_and_size(scsi_id);
/* Complete any pending commands with busy (so they get retried)
* and remove them from our queue. */
sbp2scsi_complete_all_commands(scsi_id, DID_BUS_BUSY);
/* Accept new commands unless there was another bus reset in the
* meantime. */
if (hpsb_node_entry_valid(scsi_id->ne)) {
atomic_set(&scsi_id->state, SBP2LU_STATE_RUNNING);
scsi_unblock_requests(scsi_id->scsi_host);
}
return 0;
}
static struct scsi_id_instance_data *sbp2_alloc_device(struct unit_directory *ud)
{
struct sbp2scsi_host_info *hi;
struct Scsi_Host *scsi_host = NULL;
struct scsi_id_instance_data *scsi_id = NULL;
scsi_id = kzalloc(sizeof(*scsi_id), GFP_KERNEL);
if (!scsi_id) {
SBP2_ERR("failed to create scsi_id");
goto failed_alloc;
}
scsi_id->ne = ud->ne;
scsi_id->ud = ud;
scsi_id->speed_code = IEEE1394_SPEED_100;
scsi_id->max_payload_size = sbp2_speedto_max_payload[IEEE1394_SPEED_100];
scsi_id->status_fifo_addr = CSR1212_INVALID_ADDR_SPACE;
INIT_LIST_HEAD(&scsi_id->sbp2_command_orb_inuse);
INIT_LIST_HEAD(&scsi_id->sbp2_command_orb_completed);
INIT_LIST_HEAD(&scsi_id->scsi_list);
spin_lock_init(&scsi_id->sbp2_command_orb_lock);
atomic_set(&scsi_id->state, SBP2LU_STATE_RUNNING);
INIT_WORK(&scsi_id->protocol_work, NULL);
ud->device.driver_data = scsi_id;
hi = hpsb_get_hostinfo(&sbp2_highlevel, ud->ne->host);
if (!hi) {
hi = hpsb_create_hostinfo(&sbp2_highlevel, ud->ne->host, sizeof(*hi));
if (!hi) {
SBP2_ERR("failed to allocate hostinfo");
goto failed_alloc;
}
hi->host = ud->ne->host;
INIT_LIST_HEAD(&hi->scsi_ids);
#ifdef CONFIG_IEEE1394_SBP2_PHYS_DMA
/* Handle data movement if physical dma is not
* enabled or not supported on host controller */
if (!hpsb_register_addrspace(&sbp2_highlevel, ud->ne->host,
&sbp2_physdma_ops,
0x0ULL, 0xfffffffcULL)) {
SBP2_ERR("failed to register lower 4GB address range");
goto failed_alloc;
}
#endif
}
/* Prevent unloading of the 1394 host */
if (!try_module_get(hi->host->driver->owner)) {
SBP2_ERR("failed to get a reference on 1394 host driver");
goto failed_alloc;
}
scsi_id->hi = hi;
list_add_tail(&scsi_id->scsi_list, &hi->scsi_ids);
/* Register the status FIFO address range. We could use the same FIFO
* for targets at different nodes. However we need different FIFOs per
* target in order to support multi-unit devices.
* The FIFO is located out of the local host controller's physical range
* but, if possible, within the posted write area. Status writes will
* then be performed as unified transactions. This slightly reduces
* bandwidth usage, and some Prolific based devices seem to require it.
*/
scsi_id->status_fifo_addr = hpsb_allocate_and_register_addrspace(
&sbp2_highlevel, ud->ne->host, &sbp2_ops,
sizeof(struct sbp2_status_block), sizeof(quadlet_t),
ud->ne->host->low_addr_space, CSR1212_ALL_SPACE_END);
if (scsi_id->status_fifo_addr == CSR1212_INVALID_ADDR_SPACE) {
SBP2_ERR("failed to allocate status FIFO address range");
goto failed_alloc;
}
scsi_host = scsi_host_alloc(&scsi_driver_template,
sizeof(unsigned long));
if (!scsi_host) {
SBP2_ERR("failed to register scsi host");
goto failed_alloc;
}
scsi_host->hostdata[0] = (unsigned long)scsi_id;
if (!scsi_add_host(scsi_host, &ud->device)) {
scsi_id->scsi_host = scsi_host;
return scsi_id;
}
SBP2_ERR("failed to add scsi host");
scsi_host_put(scsi_host);
failed_alloc:
sbp2_remove_device(scsi_id);
return NULL;
}
static void sbp2_host_reset(struct hpsb_host *host)
{
struct sbp2scsi_host_info *hi;
struct scsi_id_instance_data *scsi_id;
hi = hpsb_get_hostinfo(&sbp2_highlevel, host);
if (!hi)
return;
list_for_each_entry(scsi_id, &hi->scsi_ids, scsi_list)
if (likely(atomic_read(&scsi_id->state) !=
SBP2LU_STATE_IN_SHUTDOWN)) {
atomic_set(&scsi_id->state, SBP2LU_STATE_IN_RESET);
scsi_block_requests(scsi_id->scsi_host);
}
}
static int sbp2_start_device(struct scsi_id_instance_data *scsi_id)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
int error;
scsi_id->login_response =
pci_alloc_consistent(hi->host->pdev,
sizeof(struct sbp2_login_response),
&scsi_id->login_response_dma);
if (!scsi_id->login_response)
goto alloc_fail;
scsi_id->query_logins_orb =
pci_alloc_consistent(hi->host->pdev,
sizeof(struct sbp2_query_logins_orb),
&scsi_id->query_logins_orb_dma);
if (!scsi_id->query_logins_orb)
goto alloc_fail;
scsi_id->query_logins_response =
pci_alloc_consistent(hi->host->pdev,
sizeof(struct sbp2_query_logins_response),
&scsi_id->query_logins_response_dma);
if (!scsi_id->query_logins_response)
goto alloc_fail;
scsi_id->reconnect_orb =
pci_alloc_consistent(hi->host->pdev,
sizeof(struct sbp2_reconnect_orb),
&scsi_id->reconnect_orb_dma);
if (!scsi_id->reconnect_orb)
goto alloc_fail;
scsi_id->logout_orb =
pci_alloc_consistent(hi->host->pdev,
sizeof(struct sbp2_logout_orb),
&scsi_id->logout_orb_dma);
if (!scsi_id->logout_orb)
goto alloc_fail;
scsi_id->login_orb =
pci_alloc_consistent(hi->host->pdev,
sizeof(struct sbp2_login_orb),
&scsi_id->login_orb_dma);
if (!scsi_id->login_orb)
goto alloc_fail;
if (sbp2util_create_command_orb_pool(scsi_id)) {
SBP2_ERR("sbp2util_create_command_orb_pool failed!");
sbp2_remove_device(scsi_id);
return -ENOMEM;
}
/* Wait a second before trying to log in. Previously logged in
* initiators need a chance to reconnect. */
if (msleep_interruptible(1000)) {
sbp2_remove_device(scsi_id);
return -EINTR;
}
if (sbp2_login_device(scsi_id)) {
sbp2_remove_device(scsi_id);
return -EBUSY;
}
sbp2_set_busy_timeout(scsi_id);
sbp2_agent_reset(scsi_id, 1);
sbp2_max_speed_and_size(scsi_id);
error = scsi_add_device(scsi_id->scsi_host, 0, scsi_id->ud->id, 0);
if (error) {
SBP2_ERR("scsi_add_device failed");
sbp2_logout_device(scsi_id);
sbp2_remove_device(scsi_id);
return error;
}
return 0;
alloc_fail:
SBP2_ERR("Could not allocate memory for scsi_id");
sbp2_remove_device(scsi_id);
return -ENOMEM;
}
static void sbp2_remove_device(struct scsi_id_instance_data *scsi_id)
{
struct sbp2scsi_host_info *hi;
if (!scsi_id)
return;
hi = scsi_id->hi;
if (scsi_id->scsi_host) {
scsi_remove_host(scsi_id->scsi_host);
scsi_host_put(scsi_id->scsi_host);
}
flush_scheduled_work();
sbp2util_remove_command_orb_pool(scsi_id);
list_del(&scsi_id->scsi_list);
if (scsi_id->login_response)
pci_free_consistent(hi->host->pdev,
sizeof(struct sbp2_login_response),
scsi_id->login_response,
scsi_id->login_response_dma);
if (scsi_id->login_orb)
pci_free_consistent(hi->host->pdev,
sizeof(struct sbp2_login_orb),
scsi_id->login_orb,
scsi_id->login_orb_dma);
if (scsi_id->reconnect_orb)
pci_free_consistent(hi->host->pdev,
sizeof(struct sbp2_reconnect_orb),
scsi_id->reconnect_orb,
scsi_id->reconnect_orb_dma);
if (scsi_id->logout_orb)
pci_free_consistent(hi->host->pdev,
sizeof(struct sbp2_logout_orb),
scsi_id->logout_orb,
scsi_id->logout_orb_dma);
if (scsi_id->query_logins_orb)
pci_free_consistent(hi->host->pdev,
sizeof(struct sbp2_query_logins_orb),
scsi_id->query_logins_orb,
scsi_id->query_logins_orb_dma);
if (scsi_id->query_logins_response)
pci_free_consistent(hi->host->pdev,
sizeof(struct sbp2_query_logins_response),
scsi_id->query_logins_response,
scsi_id->query_logins_response_dma);
if (scsi_id->status_fifo_addr != CSR1212_INVALID_ADDR_SPACE)
hpsb_unregister_addrspace(&sbp2_highlevel, hi->host,
scsi_id->status_fifo_addr);
scsi_id->ud->device.driver_data = NULL;
if (hi)
module_put(hi->host->driver->owner);
kfree(scsi_id);
}
#ifdef CONFIG_IEEE1394_SBP2_PHYS_DMA
/*
* Deal with write requests on adapters which do not support physical DMA or
* have it switched off.
*/
static int sbp2_handle_physdma_write(struct hpsb_host *host, int nodeid,
int destid, quadlet_t *data, u64 addr,
size_t length, u16 flags)
{
memcpy(bus_to_virt((u32) addr), data, length);
return RCODE_COMPLETE;
}
/*
* Deal with read requests on adapters which do not support physical DMA or
* have it switched off.
*/
static int sbp2_handle_physdma_read(struct hpsb_host *host, int nodeid,
quadlet_t *data, u64 addr, size_t length,
u16 flags)
{
memcpy(data, bus_to_virt((u32) addr), length);
return RCODE_COMPLETE;
}
#endif
/**************************************
* SBP-2 protocol related section
**************************************/
static int sbp2_query_logins(struct scsi_id_instance_data *scsi_id)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
quadlet_t data[2];
int max_logins;
int active_logins;
scsi_id->query_logins_orb->reserved1 = 0x0;
scsi_id->query_logins_orb->reserved2 = 0x0;
scsi_id->query_logins_orb->query_response_lo = scsi_id->query_logins_response_dma;
scsi_id->query_logins_orb->query_response_hi = ORB_SET_NODE_ID(hi->host->node_id);
scsi_id->query_logins_orb->lun_misc = ORB_SET_FUNCTION(SBP2_QUERY_LOGINS_REQUEST);
scsi_id->query_logins_orb->lun_misc |= ORB_SET_NOTIFY(1);
scsi_id->query_logins_orb->lun_misc |= ORB_SET_LUN(scsi_id->sbp2_lun);
scsi_id->query_logins_orb->reserved_resp_length =
ORB_SET_QUERY_LOGINS_RESP_LENGTH(sizeof(struct sbp2_query_logins_response));
scsi_id->query_logins_orb->status_fifo_hi =
ORB_SET_STATUS_FIFO_HI(scsi_id->status_fifo_addr, hi->host->node_id);
scsi_id->query_logins_orb->status_fifo_lo =
ORB_SET_STATUS_FIFO_LO(scsi_id->status_fifo_addr);
sbp2util_cpu_to_be32_buffer(scsi_id->query_logins_orb, sizeof(struct sbp2_query_logins_orb));
memset(scsi_id->query_logins_response, 0, sizeof(struct sbp2_query_logins_response));
data[0] = ORB_SET_NODE_ID(hi->host->node_id);
data[1] = scsi_id->query_logins_orb_dma;
sbp2util_cpu_to_be32_buffer(data, 8);
hpsb_node_write(scsi_id->ne, scsi_id->sbp2_management_agent_addr, data, 8);
if (sbp2util_access_timeout(scsi_id, 2*HZ)) {
SBP2_INFO("Error querying logins to SBP-2 device - timed out");
return -EIO;
}
if (scsi_id->status_block.ORB_offset_lo != scsi_id->query_logins_orb_dma) {
SBP2_INFO("Error querying logins to SBP-2 device - timed out");
return -EIO;
}
if (STATUS_TEST_RDS(scsi_id->status_block.ORB_offset_hi_misc)) {
SBP2_INFO("Error querying logins to SBP-2 device - failed");
return -EIO;
}
sbp2util_cpu_to_be32_buffer(scsi_id->query_logins_response, sizeof(struct sbp2_query_logins_response));
max_logins = RESPONSE_GET_MAX_LOGINS(scsi_id->query_logins_response->length_max_logins);
SBP2_INFO("Maximum concurrent logins supported: %d", max_logins);
active_logins = RESPONSE_GET_ACTIVE_LOGINS(scsi_id->query_logins_response->length_max_logins);
SBP2_INFO("Number of active logins: %d", active_logins);
if (active_logins >= max_logins) {
return -EIO;
}
return 0;
}
static int sbp2_login_device(struct scsi_id_instance_data *scsi_id)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
quadlet_t data[2];
if (!scsi_id->login_orb)
return -EIO;
if (!exclusive_login) {
if (sbp2_query_logins(scsi_id)) {
SBP2_INFO("Device does not support any more concurrent logins");
return -EIO;
}
}
/* assume no password */
scsi_id->login_orb->password_hi = 0;
scsi_id->login_orb->password_lo = 0;
scsi_id->login_orb->login_response_lo = scsi_id->login_response_dma;
scsi_id->login_orb->login_response_hi = ORB_SET_NODE_ID(hi->host->node_id);
scsi_id->login_orb->lun_misc = ORB_SET_FUNCTION(SBP2_LOGIN_REQUEST);
/* one second reconnect time */
scsi_id->login_orb->lun_misc |= ORB_SET_RECONNECT(0);
scsi_id->login_orb->lun_misc |= ORB_SET_EXCLUSIVE(exclusive_login);
scsi_id->login_orb->lun_misc |= ORB_SET_NOTIFY(1);
scsi_id->login_orb->lun_misc |= ORB_SET_LUN(scsi_id->sbp2_lun);
scsi_id->login_orb->passwd_resp_lengths =
ORB_SET_LOGIN_RESP_LENGTH(sizeof(struct sbp2_login_response));
scsi_id->login_orb->status_fifo_hi =
ORB_SET_STATUS_FIFO_HI(scsi_id->status_fifo_addr, hi->host->node_id);
scsi_id->login_orb->status_fifo_lo =
ORB_SET_STATUS_FIFO_LO(scsi_id->status_fifo_addr);
sbp2util_cpu_to_be32_buffer(scsi_id->login_orb, sizeof(struct sbp2_login_orb));
memset(scsi_id->login_response, 0, sizeof(struct sbp2_login_response));
data[0] = ORB_SET_NODE_ID(hi->host->node_id);
data[1] = scsi_id->login_orb_dma;
sbp2util_cpu_to_be32_buffer(data, 8);
hpsb_node_write(scsi_id->ne, scsi_id->sbp2_management_agent_addr, data, 8);
/* wait up to 20 seconds for login status */
if (sbp2util_access_timeout(scsi_id, 20*HZ)) {
SBP2_ERR("Error logging into SBP-2 device - timed out");
return -EIO;
}
/* make sure that the returned status matches the login ORB */
if (scsi_id->status_block.ORB_offset_lo != scsi_id->login_orb_dma) {
SBP2_ERR("Error logging into SBP-2 device - timed out");
return -EIO;
}
if (STATUS_TEST_RDS(scsi_id->status_block.ORB_offset_hi_misc)) {
SBP2_ERR("Error logging into SBP-2 device - failed");
return -EIO;
}
sbp2util_cpu_to_be32_buffer(scsi_id->login_response, sizeof(struct sbp2_login_response));
scsi_id->sbp2_command_block_agent_addr =
((u64)scsi_id->login_response->command_block_agent_hi) << 32;
scsi_id->sbp2_command_block_agent_addr |= ((u64)scsi_id->login_response->command_block_agent_lo);
scsi_id->sbp2_command_block_agent_addr &= 0x0000ffffffffffffULL;
SBP2_INFO("Logged into SBP-2 device");
return 0;
}
static int sbp2_logout_device(struct scsi_id_instance_data *scsi_id)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
quadlet_t data[2];
int error;
scsi_id->logout_orb->reserved1 = 0x0;
scsi_id->logout_orb->reserved2 = 0x0;
scsi_id->logout_orb->reserved3 = 0x0;
scsi_id->logout_orb->reserved4 = 0x0;
scsi_id->logout_orb->login_ID_misc = ORB_SET_FUNCTION(SBP2_LOGOUT_REQUEST);
scsi_id->logout_orb->login_ID_misc |= ORB_SET_LOGIN_ID(scsi_id->login_response->length_login_ID);
scsi_id->logout_orb->login_ID_misc |= ORB_SET_NOTIFY(1);
scsi_id->logout_orb->reserved5 = 0x0;
scsi_id->logout_orb->status_fifo_hi =
ORB_SET_STATUS_FIFO_HI(scsi_id->status_fifo_addr, hi->host->node_id);
scsi_id->logout_orb->status_fifo_lo =
ORB_SET_STATUS_FIFO_LO(scsi_id->status_fifo_addr);
sbp2util_cpu_to_be32_buffer(scsi_id->logout_orb, sizeof(struct sbp2_logout_orb));
data[0] = ORB_SET_NODE_ID(hi->host->node_id);
data[1] = scsi_id->logout_orb_dma;
sbp2util_cpu_to_be32_buffer(data, 8);
error = hpsb_node_write(scsi_id->ne,
scsi_id->sbp2_management_agent_addr, data, 8);
if (error)
return error;
/* wait up to 1 second for the device to complete logout */
if (sbp2util_access_timeout(scsi_id, HZ))
return -EIO;
SBP2_INFO("Logged out of SBP-2 device");
return 0;
}
static int sbp2_reconnect_device(struct scsi_id_instance_data *scsi_id)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
quadlet_t data[2];
int error;
scsi_id->reconnect_orb->reserved1 = 0x0;
scsi_id->reconnect_orb->reserved2 = 0x0;
scsi_id->reconnect_orb->reserved3 = 0x0;
scsi_id->reconnect_orb->reserved4 = 0x0;
scsi_id->reconnect_orb->login_ID_misc = ORB_SET_FUNCTION(SBP2_RECONNECT_REQUEST);
scsi_id->reconnect_orb->login_ID_misc |=
ORB_SET_LOGIN_ID(scsi_id->login_response->length_login_ID);
scsi_id->reconnect_orb->login_ID_misc |= ORB_SET_NOTIFY(1);
scsi_id->reconnect_orb->reserved5 = 0x0;
scsi_id->reconnect_orb->status_fifo_hi =
ORB_SET_STATUS_FIFO_HI(scsi_id->status_fifo_addr, hi->host->node_id);
scsi_id->reconnect_orb->status_fifo_lo =
ORB_SET_STATUS_FIFO_LO(scsi_id->status_fifo_addr);
sbp2util_cpu_to_be32_buffer(scsi_id->reconnect_orb, sizeof(struct sbp2_reconnect_orb));
data[0] = ORB_SET_NODE_ID(hi->host->node_id);
data[1] = scsi_id->reconnect_orb_dma;
sbp2util_cpu_to_be32_buffer(data, 8);
error = hpsb_node_write(scsi_id->ne,
scsi_id->sbp2_management_agent_addr, data, 8);
if (error)
return error;
/* wait up to 1 second for reconnect status */
if (sbp2util_access_timeout(scsi_id, HZ)) {
SBP2_ERR("Error reconnecting to SBP-2 device - timed out");
return -EIO;
}
/* make sure that the returned status matches the reconnect ORB */
if (scsi_id->status_block.ORB_offset_lo != scsi_id->reconnect_orb_dma) {
SBP2_ERR("Error reconnecting to SBP-2 device - timed out");
return -EIO;
}
if (STATUS_TEST_RDS(scsi_id->status_block.ORB_offset_hi_misc)) {
SBP2_ERR("Error reconnecting to SBP-2 device - failed");
return -EIO;
}
SBP2_INFO("Reconnected to SBP-2 device");
return 0;
}
/*
* Set the target node's Single Phase Retry limit. Affects the target's retry
* behaviour if our node is too busy to accept requests.
*/
static int sbp2_set_busy_timeout(struct scsi_id_instance_data *scsi_id)
{
quadlet_t data;
data = cpu_to_be32(SBP2_BUSY_TIMEOUT_VALUE);
if (hpsb_node_write(scsi_id->ne, SBP2_BUSY_TIMEOUT_ADDRESS, &data, 4))
SBP2_ERR("%s error", __FUNCTION__);
return 0;
}
static void sbp2_parse_unit_directory(struct scsi_id_instance_data *scsi_id,
struct unit_directory *ud)
{
struct csr1212_keyval *kv;
struct csr1212_dentry *dentry;
u64 management_agent_addr;
u32 command_set_spec_id, command_set, unit_characteristics,
firmware_revision;
unsigned workarounds;
int i;
management_agent_addr = 0x0;
command_set_spec_id = 0x0;
command_set = 0x0;
unit_characteristics = 0x0;
firmware_revision = 0x0;
csr1212_for_each_dir_entry(ud->ne->csr, kv, ud->ud_kv, dentry) {
switch (kv->key.id) {
case CSR1212_KV_ID_DEPENDENT_INFO:
if (kv->key.type == CSR1212_KV_TYPE_CSR_OFFSET)
management_agent_addr =
CSR1212_REGISTER_SPACE_BASE +
(kv->value.csr_offset << 2);
else if (kv->key.type == CSR1212_KV_TYPE_IMMEDIATE)
scsi_id->sbp2_lun =
ORB_SET_LUN(kv->value.immediate);
break;
case SBP2_COMMAND_SET_SPEC_ID_KEY:
command_set_spec_id = kv->value.immediate;
break;
case SBP2_COMMAND_SET_KEY:
command_set = kv->value.immediate;
break;
case SBP2_UNIT_CHARACTERISTICS_KEY:
/* FIXME: This is ignored so far.
* See SBP-2 clause 7.4.8. */
unit_characteristics = kv->value.immediate;
break;
case SBP2_FIRMWARE_REVISION_KEY:
firmware_revision = kv->value.immediate;
break;
default:
/* FIXME: Check for SBP2_DEVICE_TYPE_AND_LUN_KEY.
* Its "ordered" bit has consequences for command ORB
* list handling. See SBP-2 clauses 4.6, 7.4.11, 10.2 */
break;
}
}
workarounds = sbp2_default_workarounds;
if (!(workarounds & SBP2_WORKAROUND_OVERRIDE))
for (i = 0; i < ARRAY_SIZE(sbp2_workarounds_table); i++) {
if (sbp2_workarounds_table[i].firmware_revision &&
sbp2_workarounds_table[i].firmware_revision !=
(firmware_revision & 0xffff00))
continue;
if (sbp2_workarounds_table[i].model_id &&
sbp2_workarounds_table[i].model_id != ud->model_id)
continue;
workarounds |= sbp2_workarounds_table[i].workarounds;
break;
}
if (workarounds)
SBP2_INFO("Workarounds for node " NODE_BUS_FMT ": 0x%x "
"(firmware_revision 0x%06x, vendor_id 0x%06x,"
" model_id 0x%06x)",
NODE_BUS_ARGS(ud->ne->host, ud->ne->nodeid),
workarounds, firmware_revision,
ud->vendor_id ? ud->vendor_id : ud->ne->vendor_id,
ud->model_id);
/* We would need one SCSI host template for each target to adjust
* max_sectors on the fly, therefore warn only. */
if (workarounds & SBP2_WORKAROUND_128K_MAX_TRANS &&
(max_sectors * 512) > (128 * 1024))
SBP2_INFO("Node " NODE_BUS_FMT ": Bridge only supports 128KB "
"max transfer size. WARNING: Current max_sectors "
"setting is larger than 128KB (%d sectors)",
NODE_BUS_ARGS(ud->ne->host, ud->ne->nodeid),
max_sectors);
/* If this is a logical unit directory entry, process the parent
* to get the values. */
if (ud->flags & UNIT_DIRECTORY_LUN_DIRECTORY) {
struct unit_directory *parent_ud =
container_of(ud->device.parent, struct unit_directory, device);
sbp2_parse_unit_directory(scsi_id, parent_ud);
} else {
scsi_id->sbp2_management_agent_addr = management_agent_addr;
scsi_id->sbp2_command_set_spec_id = command_set_spec_id;
scsi_id->sbp2_command_set = command_set;
scsi_id->sbp2_unit_characteristics = unit_characteristics;
scsi_id->sbp2_firmware_revision = firmware_revision;
scsi_id->workarounds = workarounds;
if (ud->flags & UNIT_DIRECTORY_HAS_LUN)
scsi_id->sbp2_lun = ORB_SET_LUN(ud->lun);
}
}
#define SBP2_PAYLOAD_TO_BYTES(p) (1 << ((p) + 2))
/*
* This function is called in order to determine the max speed and packet
* size we can use in our ORBs. Note, that we (the driver and host) only
* initiate the transaction. The SBP-2 device actually transfers the data
* (by reading from the DMA area we tell it). This means that the SBP-2
* device decides the actual maximum data it can transfer. We just tell it
* the speed that it needs to use, and the max_rec the host supports, and
* it takes care of the rest.
*/
static int sbp2_max_speed_and_size(struct scsi_id_instance_data *scsi_id)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
u8 payload;
scsi_id->speed_code =
hi->host->speed[NODEID_TO_NODE(scsi_id->ne->nodeid)];
if (scsi_id->speed_code > max_speed) {
scsi_id->speed_code = max_speed;
SBP2_INFO("Reducing speed to %s", hpsb_speedto_str[max_speed]);
}
/* Payload size is the lesser of what our speed supports and what
* our host supports. */
payload = min(sbp2_speedto_max_payload[scsi_id->speed_code],
(u8) (hi->host->csr.max_rec - 1));
/* If physical DMA is off, work around limitation in ohci1394:
* packet size must not exceed PAGE_SIZE */
if (scsi_id->ne->host->low_addr_space < (1ULL << 32))
while (SBP2_PAYLOAD_TO_BYTES(payload) + 24 > PAGE_SIZE &&
payload)
payload--;
SBP2_INFO("Node " NODE_BUS_FMT ": Max speed [%s] - Max payload [%u]",
NODE_BUS_ARGS(hi->host, scsi_id->ne->nodeid),
hpsb_speedto_str[scsi_id->speed_code],
SBP2_PAYLOAD_TO_BYTES(payload));
scsi_id->max_payload_size = payload;
return 0;
}
static int sbp2_agent_reset(struct scsi_id_instance_data *scsi_id, int wait)
{
quadlet_t data;
u64 addr;
int retval;
unsigned long flags;
/* cancel_delayed_work(&scsi_id->protocol_work); */
if (wait)
flush_scheduled_work();
data = ntohl(SBP2_AGENT_RESET_DATA);
addr = scsi_id->sbp2_command_block_agent_addr + SBP2_AGENT_RESET_OFFSET;
if (wait)
retval = hpsb_node_write(scsi_id->ne, addr, &data, 4);
else
retval = sbp2util_node_write_no_wait(scsi_id->ne, addr, &data, 4);
if (retval < 0) {
SBP2_ERR("hpsb_node_write failed.\n");
return -EIO;
}
/* make sure that the ORB_POINTER is written on next command */
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
scsi_id->last_orb = NULL;
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
return 0;
}
static void sbp2_prep_command_orb_sg(struct sbp2_command_orb *orb,
struct sbp2scsi_host_info *hi,
struct sbp2_command_info *command,
unsigned int scsi_use_sg,
struct scatterlist *sgpnt,
u32 orb_direction,
enum dma_data_direction dma_dir)
{
command->dma_dir = dma_dir;
orb->data_descriptor_hi = ORB_SET_NODE_ID(hi->host->node_id);
orb->misc |= ORB_SET_DIRECTION(orb_direction);
/* special case if only one element (and less than 64KB in size) */
if ((scsi_use_sg == 1) &&
(sgpnt[0].length <= SBP2_MAX_SG_ELEMENT_LENGTH)) {
command->dma_size = sgpnt[0].length;
command->dma_type = CMD_DMA_PAGE;
command->cmd_dma = pci_map_page(hi->host->pdev,
sgpnt[0].page,
sgpnt[0].offset,
command->dma_size,
command->dma_dir);
orb->data_descriptor_lo = command->cmd_dma;
orb->misc |= ORB_SET_DATA_SIZE(command->dma_size);
} else {
struct sbp2_unrestricted_page_table *sg_element =
&command->scatter_gather_element[0];
u32 sg_count, sg_len;
dma_addr_t sg_addr;
int i, count = pci_map_sg(hi->host->pdev, sgpnt, scsi_use_sg,
dma_dir);
command->dma_size = scsi_use_sg;
command->sge_buffer = sgpnt;
/* use page tables (s/g) */
orb->misc |= ORB_SET_PAGE_TABLE_PRESENT(0x1);
orb->data_descriptor_lo = command->sge_dma;
/* loop through and fill out our SBP-2 page tables
* (and split up anything too large) */
for (i = 0, sg_count = 0 ; i < count; i++, sgpnt++) {
sg_len = sg_dma_len(sgpnt);
sg_addr = sg_dma_address(sgpnt);
while (sg_len) {
sg_element[sg_count].segment_base_lo = sg_addr;
if (sg_len > SBP2_MAX_SG_ELEMENT_LENGTH) {
sg_element[sg_count].length_segment_base_hi =
PAGE_TABLE_SET_SEGMENT_LENGTH(SBP2_MAX_SG_ELEMENT_LENGTH);
sg_addr += SBP2_MAX_SG_ELEMENT_LENGTH;
sg_len -= SBP2_MAX_SG_ELEMENT_LENGTH;
} else {
sg_element[sg_count].length_segment_base_hi =
PAGE_TABLE_SET_SEGMENT_LENGTH(sg_len);
sg_len = 0;
}
sg_count++;
}
}
orb->misc |= ORB_SET_DATA_SIZE(sg_count);
sbp2util_cpu_to_be32_buffer(sg_element,
(sizeof(struct sbp2_unrestricted_page_table)) *
sg_count);
}
}
static void sbp2_prep_command_orb_no_sg(struct sbp2_command_orb *orb,
struct sbp2scsi_host_info *hi,
struct sbp2_command_info *command,
struct scatterlist *sgpnt,
u32 orb_direction,
unsigned int scsi_request_bufflen,
void *scsi_request_buffer,
enum dma_data_direction dma_dir)
{
command->dma_dir = dma_dir;
command->dma_size = scsi_request_bufflen;
command->dma_type = CMD_DMA_SINGLE;
command->cmd_dma = pci_map_single(hi->host->pdev, scsi_request_buffer,
command->dma_size, command->dma_dir);
orb->data_descriptor_hi = ORB_SET_NODE_ID(hi->host->node_id);
orb->misc |= ORB_SET_DIRECTION(orb_direction);
/* handle case where we get a command w/o s/g enabled
* (but check for transfers larger than 64K) */
if (scsi_request_bufflen <= SBP2_MAX_SG_ELEMENT_LENGTH) {
orb->data_descriptor_lo = command->cmd_dma;
orb->misc |= ORB_SET_DATA_SIZE(scsi_request_bufflen);
} else {
/* The buffer is too large. Turn this into page tables. */
struct sbp2_unrestricted_page_table *sg_element =
&command->scatter_gather_element[0];
u32 sg_count, sg_len;
dma_addr_t sg_addr;
orb->data_descriptor_lo = command->sge_dma;
orb->misc |= ORB_SET_PAGE_TABLE_PRESENT(0x1);
/* fill out our SBP-2 page tables; split up the large buffer */
sg_count = 0;
sg_len = scsi_request_bufflen;
sg_addr = command->cmd_dma;
while (sg_len) {
sg_element[sg_count].segment_base_lo = sg_addr;
if (sg_len > SBP2_MAX_SG_ELEMENT_LENGTH) {
sg_element[sg_count].length_segment_base_hi =
PAGE_TABLE_SET_SEGMENT_LENGTH(SBP2_MAX_SG_ELEMENT_LENGTH);
sg_addr += SBP2_MAX_SG_ELEMENT_LENGTH;
sg_len -= SBP2_MAX_SG_ELEMENT_LENGTH;
} else {
sg_element[sg_count].length_segment_base_hi =
PAGE_TABLE_SET_SEGMENT_LENGTH(sg_len);
sg_len = 0;
}
sg_count++;
}
orb->misc |= ORB_SET_DATA_SIZE(sg_count);
sbp2util_cpu_to_be32_buffer(sg_element,
(sizeof(struct sbp2_unrestricted_page_table)) *
sg_count);
}
}
static void sbp2_create_command_orb(struct scsi_id_instance_data *scsi_id,
struct sbp2_command_info *command,
unchar *scsi_cmd,
unsigned int scsi_use_sg,
unsigned int scsi_request_bufflen,
void *scsi_request_buffer,
enum dma_data_direction dma_dir)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
struct scatterlist *sgpnt = (struct scatterlist *)scsi_request_buffer;
struct sbp2_command_orb *command_orb = &command->command_orb;
u32 orb_direction;
/*
* Set-up our command ORB.
*
* NOTE: We're doing unrestricted page tables (s/g), as this is
* best performance (at least with the devices I have). This means
* that data_size becomes the number of s/g elements, and
* page_size should be zero (for unrestricted).
*/
command_orb->next_ORB_hi = ORB_SET_NULL_PTR(1);
command_orb->next_ORB_lo = 0x0;
command_orb->misc = ORB_SET_MAX_PAYLOAD(scsi_id->max_payload_size);
command_orb->misc |= ORB_SET_SPEED(scsi_id->speed_code);
command_orb->misc |= ORB_SET_NOTIFY(1);
if (dma_dir == DMA_NONE)
orb_direction = ORB_DIRECTION_NO_DATA_TRANSFER;
else if (dma_dir == DMA_TO_DEVICE && scsi_request_bufflen)
orb_direction = ORB_DIRECTION_WRITE_TO_MEDIA;
else if (dma_dir == DMA_FROM_DEVICE && scsi_request_bufflen)
orb_direction = ORB_DIRECTION_READ_FROM_MEDIA;
else {
SBP2_INFO("Falling back to DMA_NONE");
orb_direction = ORB_DIRECTION_NO_DATA_TRANSFER;
}
/* set up our page table stuff */
if (orb_direction == ORB_DIRECTION_NO_DATA_TRANSFER) {
command_orb->data_descriptor_hi = 0x0;
command_orb->data_descriptor_lo = 0x0;
command_orb->misc |= ORB_SET_DIRECTION(1);
} else if (scsi_use_sg)
sbp2_prep_command_orb_sg(command_orb, hi, command, scsi_use_sg,
sgpnt, orb_direction, dma_dir);
else
sbp2_prep_command_orb_no_sg(command_orb, hi, command, sgpnt,
orb_direction, scsi_request_bufflen,
scsi_request_buffer, dma_dir);
sbp2util_cpu_to_be32_buffer(command_orb, sizeof(struct sbp2_command_orb));
memset(command_orb->cdb, 0, 12);
memcpy(command_orb->cdb, scsi_cmd, COMMAND_SIZE(*scsi_cmd));
}
static void sbp2_link_orb_command(struct scsi_id_instance_data *scsi_id,
struct sbp2_command_info *command)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
struct sbp2_command_orb *command_orb = &command->command_orb;
struct sbp2_command_orb *last_orb;
dma_addr_t last_orb_dma;
u64 addr = scsi_id->sbp2_command_block_agent_addr;
quadlet_t data[2];
size_t length;
unsigned long flags;
pci_dma_sync_single_for_device(hi->host->pdev, command->command_orb_dma,
sizeof(struct sbp2_command_orb),
PCI_DMA_TODEVICE);
pci_dma_sync_single_for_device(hi->host->pdev, command->sge_dma,
sizeof(command->scatter_gather_element),
PCI_DMA_BIDIRECTIONAL);
/* check to see if there are any previous orbs to use */
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
last_orb = scsi_id->last_orb;
last_orb_dma = scsi_id->last_orb_dma;
if (!last_orb) {
/*
* last_orb == NULL means: We know that the target's fetch agent
* is not active right now.
*/
addr += SBP2_ORB_POINTER_OFFSET;
data[0] = ORB_SET_NODE_ID(hi->host->node_id);
data[1] = command->command_orb_dma;
sbp2util_cpu_to_be32_buffer(data, 8);
length = 8;
} else {
/*
* last_orb != NULL means: We know that the target's fetch agent
* is (very probably) not dead or in reset state right now.
* We have an ORB already sent that we can append a new one to.
* The target's fetch agent may or may not have read this
* previous ORB yet.
*/
pci_dma_sync_single_for_cpu(hi->host->pdev, last_orb_dma,
sizeof(struct sbp2_command_orb),
PCI_DMA_TODEVICE);
last_orb->next_ORB_lo = cpu_to_be32(command->command_orb_dma);
wmb();
/* Tells hardware that this pointer is valid */
last_orb->next_ORB_hi = 0;
pci_dma_sync_single_for_device(hi->host->pdev, last_orb_dma,
sizeof(struct sbp2_command_orb),
PCI_DMA_TODEVICE);
addr += SBP2_DOORBELL_OFFSET;
data[0] = 0;
length = 4;
}
scsi_id->last_orb = command_orb;
scsi_id->last_orb_dma = command->command_orb_dma;
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
if (sbp2util_node_write_no_wait(scsi_id->ne, addr, data, length)) {
/*
* sbp2util_node_write_no_wait failed. We certainly ran out
* of transaction labels, perhaps just because there were no
* context switches which gave khpsbpkt a chance to collect
* free tlabels. Try again in non-atomic context. If necessary,
* the workqueue job will sleep to guaranteedly get a tlabel.
* We do not accept new commands until the job is over.
*/
scsi_block_requests(scsi_id->scsi_host);
PREPARE_WORK(&scsi_id->protocol_work,
last_orb ? sbp2util_write_doorbell:
sbp2util_write_orb_pointer
/* */);
schedule_work(&scsi_id->protocol_work);
}
}
static int sbp2_send_command(struct scsi_id_instance_data *scsi_id,
struct scsi_cmnd *SCpnt,
void (*done)(struct scsi_cmnd *))
{
unchar *cmd = (unchar *) SCpnt->cmnd;
unsigned int request_bufflen = SCpnt->request_bufflen;
struct sbp2_command_info *command;
command = sbp2util_allocate_command_orb(scsi_id, SCpnt, done);
if (!command)
return -EIO;
sbp2_create_command_orb(scsi_id, command, cmd, SCpnt->use_sg,
request_bufflen, SCpnt->request_buffer,
SCpnt->sc_data_direction);
sbp2_link_orb_command(scsi_id, command);
return 0;
}
/*
* Translates SBP-2 status into SCSI sense data for check conditions
*/
static unsigned int sbp2_status_to_sense_data(unchar *sbp2_status, unchar *sense_data)
{
/* OK, it's pretty ugly... ;-) */
sense_data[0] = 0x70;
sense_data[1] = 0x0;
sense_data[2] = sbp2_status[9];
sense_data[3] = sbp2_status[12];
sense_data[4] = sbp2_status[13];
sense_data[5] = sbp2_status[14];
sense_data[6] = sbp2_status[15];
sense_data[7] = 10;
sense_data[8] = sbp2_status[16];
sense_data[9] = sbp2_status[17];
sense_data[10] = sbp2_status[18];
sense_data[11] = sbp2_status[19];
sense_data[12] = sbp2_status[10];
sense_data[13] = sbp2_status[11];
sense_data[14] = sbp2_status[20];
sense_data[15] = sbp2_status[21];
return sbp2_status[8] & 0x3f;
}
static int sbp2_handle_status_write(struct hpsb_host *host, int nodeid,
int destid, quadlet_t *data, u64 addr,
size_t length, u16 fl)
{
struct sbp2scsi_host_info *hi;
struct scsi_id_instance_data *scsi_id = NULL, *scsi_id_tmp;
struct scsi_cmnd *SCpnt = NULL;
struct sbp2_status_block *sb;
u32 scsi_status = SBP2_SCSI_STATUS_GOOD;
struct sbp2_command_info *command;
unsigned long flags;
if (unlikely(length < 8 || length > sizeof(struct sbp2_status_block))) {
SBP2_ERR("Wrong size of status block");
return RCODE_ADDRESS_ERROR;
}
if (unlikely(!host)) {
SBP2_ERR("host is NULL - this is bad!");
return RCODE_ADDRESS_ERROR;
}
hi = hpsb_get_hostinfo(&sbp2_highlevel, host);
if (unlikely(!hi)) {
SBP2_ERR("host info is NULL - this is bad!");
return RCODE_ADDRESS_ERROR;
}
/* Find the unit which wrote the status. */
list_for_each_entry(scsi_id_tmp, &hi->scsi_ids, scsi_list) {
if (scsi_id_tmp->ne->nodeid == nodeid &&
scsi_id_tmp->status_fifo_addr == addr) {
scsi_id = scsi_id_tmp;
break;
}
}
if (unlikely(!scsi_id)) {
SBP2_ERR("scsi_id is NULL - device is gone?");
return RCODE_ADDRESS_ERROR;
}
/* Put response into scsi_id status fifo buffer. The first two bytes
* come in big endian bit order. Often the target writes only a
* truncated status block, minimally the first two quadlets. The rest
* is implied to be zeros. */
sb = &scsi_id->status_block;
memset(sb->command_set_dependent, 0, sizeof(sb->command_set_dependent));
memcpy(sb, data, length);
sbp2util_be32_to_cpu_buffer(sb, 8);
/* Ignore unsolicited status. Handle command ORB status. */
if (unlikely(STATUS_GET_SRC(sb->ORB_offset_hi_misc) == 2))
command = NULL;
else
command = sbp2util_find_command_for_orb(scsi_id,
sb->ORB_offset_lo);
if (command) {
pci_dma_sync_single_for_cpu(hi->host->pdev, command->command_orb_dma,
sizeof(struct sbp2_command_orb),
PCI_DMA_TODEVICE);
pci_dma_sync_single_for_cpu(hi->host->pdev, command->sge_dma,
sizeof(command->scatter_gather_element),
PCI_DMA_BIDIRECTIONAL);
/* Grab SCSI command pointers and check status. */
/*
* FIXME: If the src field in the status is 1, the ORB DMA must
* not be reused until status for a subsequent ORB is received.
*/
SCpnt = command->Current_SCpnt;
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
sbp2util_mark_command_completed(scsi_id, command);
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
if (SCpnt) {
u32 h = sb->ORB_offset_hi_misc;
u32 r = STATUS_GET_RESP(h);
if (r != RESP_STATUS_REQUEST_COMPLETE) {
SBP2_INFO("resp 0x%x, sbp_status 0x%x",
r, STATUS_GET_SBP_STATUS(h));
scsi_status =
r == RESP_STATUS_TRANSPORT_FAILURE ?
SBP2_SCSI_STATUS_BUSY :
SBP2_SCSI_STATUS_COMMAND_TERMINATED;
}
if (STATUS_GET_LEN(h) > 1)
scsi_status = sbp2_status_to_sense_data(
(unchar *)sb, SCpnt->sense_buffer);
if (STATUS_TEST_DEAD(h))
sbp2_agent_reset(scsi_id, 0);
}
/* Check here to see if there are no commands in-use. If there
* are none, we know that the fetch agent left the active state
* _and_ that we did not reactivate it yet. Therefore clear
* last_orb so that next time we write directly to the
* ORB_POINTER register. That way the fetch agent does not need
* to refetch the next_ORB. */
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
if (list_empty(&scsi_id->sbp2_command_orb_inuse))
scsi_id->last_orb = NULL;
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
} else {
/* It's probably status after a management request. */
if ((sb->ORB_offset_lo == scsi_id->reconnect_orb_dma) ||
(sb->ORB_offset_lo == scsi_id->login_orb_dma) ||
(sb->ORB_offset_lo == scsi_id->query_logins_orb_dma) ||
(sb->ORB_offset_lo == scsi_id->logout_orb_dma)) {
scsi_id->access_complete = 1;
wake_up_interruptible(&access_wq);
}
}
if (SCpnt)
sbp2scsi_complete_command(scsi_id, scsi_status, SCpnt,
command->Current_done);
return RCODE_COMPLETE;
}
/**************************************
* SCSI interface related section
**************************************/
static int sbp2scsi_queuecommand(struct scsi_cmnd *SCpnt,
void (*done)(struct scsi_cmnd *))
{
struct scsi_id_instance_data *scsi_id =
(struct scsi_id_instance_data *)SCpnt->device->host->hostdata[0];
struct sbp2scsi_host_info *hi;
int result = DID_NO_CONNECT << 16;
if (unlikely(!sbp2util_node_is_available(scsi_id)))
goto done;
hi = scsi_id->hi;
if (unlikely(!hi)) {
SBP2_ERR("sbp2scsi_host_info is NULL - this is bad!");
goto done;
}
/* Multiple units are currently represented to the SCSI core as separate
* targets, not as one target with multiple LUs. Therefore return
* selection time-out to any IO directed at non-zero LUNs. */
if (unlikely(SCpnt->device->lun))
goto done;
/* handle the request sense command here (auto-request sense) */
if (SCpnt->cmnd[0] == REQUEST_SENSE) {
memcpy(SCpnt->request_buffer, SCpnt->sense_buffer, SCpnt->request_bufflen);
memset(SCpnt->sense_buffer, 0, sizeof(SCpnt->sense_buffer));
sbp2scsi_complete_command(scsi_id, SBP2_SCSI_STATUS_GOOD, SCpnt, done);
return 0;
}
if (unlikely(!hpsb_node_entry_valid(scsi_id->ne))) {
SBP2_ERR("Bus reset in progress - rejecting command");
result = DID_BUS_BUSY << 16;
goto done;
}
/* Bidirectional commands are not yet implemented,
* and unknown transfer direction not handled. */
if (unlikely(SCpnt->sc_data_direction == DMA_BIDIRECTIONAL)) {
SBP2_ERR("Cannot handle DMA_BIDIRECTIONAL - rejecting command");
result = DID_ERROR << 16;
goto done;
}
if (sbp2_send_command(scsi_id, SCpnt, done)) {
SBP2_ERR("Error sending SCSI command");
sbp2scsi_complete_command(scsi_id, SBP2_SCSI_STATUS_SELECTION_TIMEOUT,
SCpnt, done);
}
return 0;
done:
SCpnt->result = result;
done(SCpnt);
return 0;
}
static void sbp2scsi_complete_all_commands(struct scsi_id_instance_data *scsi_id,
u32 status)
{
struct sbp2scsi_host_info *hi = scsi_id->hi;
struct list_head *lh;
struct sbp2_command_info *command;
unsigned long flags;
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
while (!list_empty(&scsi_id->sbp2_command_orb_inuse)) {
lh = scsi_id->sbp2_command_orb_inuse.next;
command = list_entry(lh, struct sbp2_command_info, list);
pci_dma_sync_single_for_cpu(hi->host->pdev, command->command_orb_dma,
sizeof(struct sbp2_command_orb),
PCI_DMA_TODEVICE);
pci_dma_sync_single_for_cpu(hi->host->pdev, command->sge_dma,
sizeof(command->scatter_gather_element),
PCI_DMA_BIDIRECTIONAL);
sbp2util_mark_command_completed(scsi_id, command);
if (command->Current_SCpnt) {
command->Current_SCpnt->result = status << 16;
command->Current_done(command->Current_SCpnt);
}
}
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
return;
}
/*
* Complete a regular SCSI command. Can be called in atomic context.
*/
static void sbp2scsi_complete_command(struct scsi_id_instance_data *scsi_id,
u32 scsi_status, struct scsi_cmnd *SCpnt,
void (*done)(struct scsi_cmnd *))
{
if (!SCpnt) {
SBP2_ERR("SCpnt is NULL");
return;
}
switch (scsi_status) {
case SBP2_SCSI_STATUS_GOOD:
SCpnt->result = DID_OK << 16;
break;
case SBP2_SCSI_STATUS_BUSY:
SBP2_ERR("SBP2_SCSI_STATUS_BUSY");
SCpnt->result = DID_BUS_BUSY << 16;
break;
case SBP2_SCSI_STATUS_CHECK_CONDITION:
SCpnt->result = CHECK_CONDITION << 1 | DID_OK << 16;
break;
case SBP2_SCSI_STATUS_SELECTION_TIMEOUT:
SBP2_ERR("SBP2_SCSI_STATUS_SELECTION_TIMEOUT");
SCpnt->result = DID_NO_CONNECT << 16;
scsi_print_command(SCpnt);
break;
case SBP2_SCSI_STATUS_CONDITION_MET:
case SBP2_SCSI_STATUS_RESERVATION_CONFLICT:
case SBP2_SCSI_STATUS_COMMAND_TERMINATED:
SBP2_ERR("Bad SCSI status = %x", scsi_status);
SCpnt->result = DID_ERROR << 16;
scsi_print_command(SCpnt);
break;
default:
SBP2_ERR("Unsupported SCSI status = %x", scsi_status);
SCpnt->result = DID_ERROR << 16;
}
/* If a bus reset is in progress and there was an error, complete
* the command as busy so that it will get retried. */
if (!hpsb_node_entry_valid(scsi_id->ne)
&& (scsi_status != SBP2_SCSI_STATUS_GOOD)) {
SBP2_ERR("Completing command with busy (bus reset)");
SCpnt->result = DID_BUS_BUSY << 16;
}
/* Tell the SCSI stack that we're done with this command. */
done(SCpnt);
}
static int sbp2scsi_slave_alloc(struct scsi_device *sdev)
{
struct scsi_id_instance_data *scsi_id =
(struct scsi_id_instance_data *)sdev->host->hostdata[0];
scsi_id->sdev = sdev;
sdev->allow_restart = 1;
if (scsi_id->workarounds & SBP2_WORKAROUND_INQUIRY_36)
sdev->inquiry_len = 36;
return 0;
}
static int sbp2scsi_slave_configure(struct scsi_device *sdev)
{
struct scsi_id_instance_data *scsi_id =
(struct scsi_id_instance_data *)sdev->host->hostdata[0];
blk_queue_dma_alignment(sdev->request_queue, (512 - 1));
sdev->use_10_for_rw = 1;
if (sdev->type == TYPE_DISK &&
scsi_id->workarounds & SBP2_WORKAROUND_MODE_SENSE_8)
sdev->skip_ms_page_8 = 1;
if (scsi_id->workarounds & SBP2_WORKAROUND_FIX_CAPACITY)
sdev->fix_capacity = 1;
return 0;
}
static void sbp2scsi_slave_destroy(struct scsi_device *sdev)
{
((struct scsi_id_instance_data *)sdev->host->hostdata[0])->sdev = NULL;
return;
}
/*
* Called by scsi stack when something has really gone wrong.
* Usually called when a command has timed-out for some reason.
*/
static int sbp2scsi_abort(struct scsi_cmnd *SCpnt)
{
struct scsi_id_instance_data *scsi_id =
(struct scsi_id_instance_data *)SCpnt->device->host->hostdata[0];
struct sbp2scsi_host_info *hi = scsi_id->hi;
struct sbp2_command_info *command;
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
unsigned long flags;
SBP2_INFO("aborting sbp2 command");
scsi_print_command(SCpnt);
if (sbp2util_node_is_available(scsi_id)) {
sbp2_agent_reset(scsi_id, 1);
/* Return a matching command structure to the free pool. */
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
spin_lock_irqsave(&scsi_id->sbp2_command_orb_lock, flags);
command = sbp2util_find_command_for_SCpnt(scsi_id, SCpnt);
if (command) {
pci_dma_sync_single_for_cpu(hi->host->pdev,
command->command_orb_dma,
sizeof(struct sbp2_command_orb),
PCI_DMA_TODEVICE);
pci_dma_sync_single_for_cpu(hi->host->pdev,
command->sge_dma,
sizeof(command->scatter_gather_element),
PCI_DMA_BIDIRECTIONAL);
sbp2util_mark_command_completed(scsi_id, command);
if (command->Current_SCpnt) {
command->Current_SCpnt->result = DID_ABORT << 16;
command->Current_done(command->Current_SCpnt);
}
}
[PATCH] sbp2: fix spinlock recursion sbp2util_mark_command_completed takes a lock which was already taken by sbp2scsi_complete_all_commands. This is a regression in Linux 2.6.15. Reported by Kristian Harms at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187394 [ More complete commentary, as response to questions by Andrew: ] > This changes the call environment for all implementations of > ->Current_done(). Are they all safe to call under this lock? Short answer: Yes, trust me. ;-) Long answer: The done() callbacks are passed on to sbp2 from the SCSI stack along with each SCSI command via the queuecommand hook. The done() callback is safe to call in atomic context. So does Documentation/scsi/scsi_mid_low_api.txt say, and many if not all SCSI low-level handlers rely on this fact. So whatever this callback does, it is "self-contained" and it won't conflict with sbp2's internal ORB list handling. In particular, it won't race with the sbp2_command_orb_lock. Moreover, sbp2 already calls the done() handler with sbp2_command_orb_lock taken in sbp2scsi_complete_all_commands(). I admit this is ultimately no proof of correctness, especially since this portion of code introduced the spinlock recursion in the first place and we didn't realize it since this code's submission before 2.6.15 until now. (I have learned a lesson from this.) I stress-tested my patch on x86 uniprocessor with a preemptible SMP kernel (alas I have no SMP machine yet) and made sure that all code paths which involve the sbp2_command_orb_lock were gone through multiple times. Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-01 19:11:41 +00:00
spin_unlock_irqrestore(&scsi_id->sbp2_command_orb_lock, flags);
sbp2scsi_complete_all_commands(scsi_id, DID_BUS_BUSY);
}
return SUCCESS;
}
/*
* Called by scsi stack when something has really gone wrong.
*/
static int sbp2scsi_reset(struct scsi_cmnd *SCpnt)
{
struct scsi_id_instance_data *scsi_id =
(struct scsi_id_instance_data *)SCpnt->device->host->hostdata[0];
SBP2_INFO("reset requested");
if (sbp2util_node_is_available(scsi_id)) {
SBP2_INFO("generating sbp2 fetch agent reset");
sbp2_agent_reset(scsi_id, 1);
}
return SUCCESS;
}
static ssize_t sbp2_sysfs_ieee1394_id_show(struct device *dev,
struct device_attribute *attr,
char *buf)
{
struct scsi_device *sdev;
struct scsi_id_instance_data *scsi_id;
int lun;
if (!(sdev = to_scsi_device(dev)))
return 0;
if (!(scsi_id = (struct scsi_id_instance_data *)sdev->host->hostdata[0]))
return 0;
lun = ORB_SET_LUN(scsi_id->sbp2_lun);
return sprintf(buf, "%016Lx:%d:%d\n", (unsigned long long)scsi_id->ne->guid,
scsi_id->ud->id, lun);
}
MODULE_AUTHOR("Ben Collins <bcollins@debian.org>");
MODULE_DESCRIPTION("IEEE-1394 SBP-2 protocol driver");
MODULE_SUPPORTED_DEVICE(SBP2_DEVICE_NAME);
MODULE_LICENSE("GPL");
static int sbp2_module_init(void)
{
int ret;
if (serialize_io) {
scsi_driver_template.can_queue = 1;
scsi_driver_template.cmd_per_lun = 1;
}
if (sbp2_default_workarounds & SBP2_WORKAROUND_128K_MAX_TRANS &&
(max_sectors * 512) > (128 * 1024))
max_sectors = 128 * 1024 / 512;
scsi_driver_template.max_sectors = max_sectors;
hpsb_register_highlevel(&sbp2_highlevel);
ret = hpsb_register_protocol(&sbp2_driver);
if (ret) {
SBP2_ERR("Failed to register protocol");
hpsb_unregister_highlevel(&sbp2_highlevel);
return ret;
}
return 0;
}
static void __exit sbp2_module_exit(void)
{
hpsb_unregister_protocol(&sbp2_driver);
hpsb_unregister_highlevel(&sbp2_highlevel);
}
module_init(sbp2_module_init);
module_exit(sbp2_module_exit);