asgardius/android_kernel_motorola_sm6225
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
android_kernel_motorola_sm6225/drivers/scsi/aic7xxx/aic79xx_osm.c
be7db055dd [PATCH] remove old scsi data direction macros 
these have been wrappers for the generic dma direction bits since 2.5.x.
This patch converts the few remaining drivers and removes the macros.

Arjan noticed there's some hunk in here that shouldn't.  Updated patch
below:

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2005-04-18 13:49:58 -05:00

5017 lines
136 KiB
C
Raw Blame History

/*
* Adaptec AIC79xx device driver for Linux.
*
* $Id: //depot/aic7xxx/linux/drivers/scsi/aic7xxx/aic79xx_osm.c#171 $
*
* --------------------------------------------------------------------------
* Copyright (c) 1994-2000 Justin T. Gibbs.
* Copyright (c) 1997-1999 Doug Ledford
* Copyright (c) 2000-2003 Adaptec Inc.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions, and the following disclaimer,
* without modification.
* 2. Redistributions in binary form must reproduce at minimum a disclaimer
* substantially similar to the "NO WARRANTY" disclaimer below
* ("Disclaimer") and any redistribution must be conditioned upon
* including a substantially similar Disclaimer requirement for further
* binary redistribution.
* 3. Neither the names of the above-listed copyright holders nor the names
* of any contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* Alternatively, this software may be distributed under the terms of the
* GNU General Public License ("GPL") version 2 as published by the Free
* Software Foundation.
*
* NO WARRANTY
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* HOLDERS OR CONTRIBUTORS BE LIABLE FOR SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
* IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGES.
*/
#include "aic79xx_osm.h"
#include "aic79xx_inline.h"
#include <scsi/scsicam.h>
/*
* Include aiclib.c as part of our
* "module dependencies are hard" work around.
*/
#include "aiclib.c"
#include <linux/init.h> /* __setup */
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
#include "sd.h" /* For geometry detection */
#endif
#include <linux/mm.h> /* For fetching system memory size */
#include <linux/delay.h> /* For ssleep/msleep */
/*
* Lock protecting manipulation of the ahd softc list.
*/
spinlock_t ahd_list_spinlock;
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
/* For dynamic sglist size calculation. */
u_int ahd_linux_nseg;
#endif
/*
* Bucket size for counting good commands in between bad ones.
*/
#define AHD_LINUX_ERR_THRESH 1000
/*
* Set this to the delay in seconds after SCSI bus reset.
* Note, we honor this only for the initial bus reset.
* The scsi error recovery code performs its own bus settle
* delay handling for error recovery actions.
*/
#ifdef CONFIG_AIC79XX_RESET_DELAY_MS
#define AIC79XX_RESET_DELAY CONFIG_AIC79XX_RESET_DELAY_MS
#else
#define AIC79XX_RESET_DELAY 5000
#endif
/*
* To change the default number of tagged transactions allowed per-device,
* add a line to the lilo.conf file like:
* append="aic79xx=verbose,tag_info:{{32,32,32,32},{32,32,32,32}}"
* which will result in the first four devices on the first two
* controllers being set to a tagged queue depth of 32.
*
* The tag_commands is an array of 16 to allow for wide and twin adapters.
* Twin adapters will use indexes 0-7 for channel 0, and indexes 8-15
* for channel 1.
*/
typedef struct {
uint16_t tag_commands[16]; /* Allow for wide/twin adapters. */
} adapter_tag_info_t;
/*
* Modify this as you see fit for your system.
*
* 0 tagged queuing disabled
* 1 <= n <= 253 n == max tags ever dispatched.
*
* The driver will throttle the number of commands dispatched to a
* device if it returns queue full. For devices with a fixed maximum
* queue depth, the driver will eventually determine this depth and
* lock it in (a console message is printed to indicate that a lock
* has occurred). On some devices, queue full is returned for a temporary
* resource shortage. These devices will return queue full at varying
* depths. The driver will throttle back when the queue fulls occur and
* attempt to slowly increase the depth over time as the device recovers
* from the resource shortage.
*
* In this example, the first line will disable tagged queueing for all
* the devices on the first probed aic79xx adapter.
*
* The second line enables tagged queueing with 4 commands/LUN for IDs
* (0, 2-11, 13-15), disables tagged queueing for ID 12, and tells the
* driver to attempt to use up to 64 tags for ID 1.
*
* The third line is the same as the first line.
*
* The fourth line disables tagged queueing for devices 0 and 3. It
* enables tagged queueing for the other IDs, with 16 commands/LUN
* for IDs 1 and 4, 127 commands/LUN for ID 8, and 4 commands/LUN for
* IDs 2, 5-7, and 9-15.
*/
/*
* NOTE: The below structure is for reference only, the actual structure
* to modify in order to change things is just below this comment block.
adapter_tag_info_t aic79xx_tag_info[] =
{
{{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}},
{{4, 64, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 0, 4, 4, 4}},
{{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}},
{{0, 16, 4, 0, 16, 4, 4, 4, 127, 4, 4, 4, 4, 4, 4, 4}}
};
*/
#ifdef CONFIG_AIC79XX_CMDS_PER_DEVICE
#define AIC79XX_CMDS_PER_DEVICE CONFIG_AIC79XX_CMDS_PER_DEVICE
#else
#define AIC79XX_CMDS_PER_DEVICE AHD_MAX_QUEUE
#endif
#define AIC79XX_CONFIGED_TAG_COMMANDS { \
AIC79XX_CMDS_PER_DEVICE, AIC79XX_CMDS_PER_DEVICE, \
AIC79XX_CMDS_PER_DEVICE, AIC79XX_CMDS_PER_DEVICE, \
AIC79XX_CMDS_PER_DEVICE, AIC79XX_CMDS_PER_DEVICE, \
AIC79XX_CMDS_PER_DEVICE, AIC79XX_CMDS_PER_DEVICE, \
AIC79XX_CMDS_PER_DEVICE, AIC79XX_CMDS_PER_DEVICE, \
AIC79XX_CMDS_PER_DEVICE, AIC79XX_CMDS_PER_DEVICE, \
AIC79XX_CMDS_PER_DEVICE, AIC79XX_CMDS_PER_DEVICE, \
AIC79XX_CMDS_PER_DEVICE, AIC79XX_CMDS_PER_DEVICE \
}
/*
* By default, use the number of commands specified by
* the users kernel configuration.
*/
static adapter_tag_info_t aic79xx_tag_info[] =
{
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS},
{AIC79XX_CONFIGED_TAG_COMMANDS}
};
/*
* By default, read streaming is disabled. In theory,
* read streaming should enhance performance, but early
* U320 drive firmware actually performs slower with
* read streaming enabled.
*/
#ifdef CONFIG_AIC79XX_ENABLE_RD_STRM
#define AIC79XX_CONFIGED_RD_STRM 0xFFFF
#else
#define AIC79XX_CONFIGED_RD_STRM 0
#endif
static uint16_t aic79xx_rd_strm_info[] =
{
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM,
AIC79XX_CONFIGED_RD_STRM
};
/*
* DV option:
*
* positive value = DV Enabled
* zero = DV Disabled
* negative value = DV Default for adapter type/seeprom
*/
#ifdef CONFIG_AIC79XX_DV_SETTING
#define AIC79XX_CONFIGED_DV CONFIG_AIC79XX_DV_SETTING
#else
#define AIC79XX_CONFIGED_DV -1
#endif
static int8_t aic79xx_dv_settings[] =
{
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV,
AIC79XX_CONFIGED_DV
};
/*
* The I/O cell on the chip is very configurable in respect to its analog
* characteristics. Set the defaults here; they can be overriden with
* the proper insmod parameters.
*/
struct ahd_linux_iocell_opts
{
uint8_t precomp;
uint8_t slewrate;
uint8_t amplitude;
};
#define AIC79XX_DEFAULT_PRECOMP 0xFF
#define AIC79XX_DEFAULT_SLEWRATE 0xFF
#define AIC79XX_DEFAULT_AMPLITUDE 0xFF
#define AIC79XX_DEFAULT_IOOPTS \
{ \
AIC79XX_DEFAULT_PRECOMP, \
AIC79XX_DEFAULT_SLEWRATE, \
AIC79XX_DEFAULT_AMPLITUDE \
}
#define AIC79XX_PRECOMP_INDEX 0
#define AIC79XX_SLEWRATE_INDEX 1
#define AIC79XX_AMPLITUDE_INDEX 2
static struct ahd_linux_iocell_opts aic79xx_iocell_info[] =
{
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS,
AIC79XX_DEFAULT_IOOPTS
};
/*
* There should be a specific return value for this in scsi.h, but
* it seems that most drivers ignore it.
*/
#define DID_UNDERFLOW DID_ERROR
void
ahd_print_path(struct ahd_softc *ahd, struct scb *scb)
{
printk("(scsi%d:%c:%d:%d): ",
ahd->platform_data->host->host_no,
scb != NULL ? SCB_GET_CHANNEL(ahd, scb) : 'X',
scb != NULL ? SCB_GET_TARGET(ahd, scb) : -1,
scb != NULL ? SCB_GET_LUN(scb) : -1);
}
/*
* XXX - these options apply unilaterally to _all_ adapters
* cards in the system. This should be fixed. Exceptions to this
* rule are noted in the comments.
*/
/*
* Skip the scsi bus reset. Non 0 make us skip the reset at startup. This
* has no effect on any later resets that might occur due to things like
* SCSI bus timeouts.
*/
static uint32_t aic79xx_no_reset;
/*
* Certain PCI motherboards will scan PCI devices from highest to lowest,
* others scan from lowest to highest, and they tend to do all kinds of
* strange things when they come into contact with PCI bridge chips. The
* net result of all this is that the PCI card that is actually used to boot
* the machine is very hard to detect. Most motherboards go from lowest
* PCI slot number to highest, and the first SCSI controller found is the
* one you boot from. The only exceptions to this are when a controller
* has its BIOS disabled. So, we by default sort all of our SCSI controllers
* from lowest PCI slot number to highest PCI slot number. We also force
* all controllers with their BIOS disabled to the end of the list. This
* works on *almost* all computers. Where it doesn't work, we have this
* option. Setting this option to non-0 will reverse the order of the sort
* to highest first, then lowest, but will still leave cards with their BIOS
* disabled at the very end. That should fix everyone up unless there are
* really strange cirumstances.
*/
static uint32_t aic79xx_reverse_scan;
/*
* Should we force EXTENDED translation on a controller.
* 0 == Use whatever is in the SEEPROM or default to off
* 1 == Use whatever is in the SEEPROM or default to on
*/
static uint32_t aic79xx_extended;
/*
* PCI bus parity checking of the Adaptec controllers. This is somewhat
* dubious at best. To my knowledge, this option has never actually
* solved a PCI parity problem, but on certain machines with broken PCI
* chipset configurations, it can generate tons of false error messages.
* It's included in the driver for completeness.
* 0 = Shut off PCI parity check
* non-0 = Enable PCI parity check
*
* NOTE: you can't actually pass -1 on the lilo prompt. So, to set this
* variable to -1 you would actually want to simply pass the variable
* name without a number. That will invert the 0 which will result in
* -1.
*/
static uint32_t aic79xx_pci_parity = ~0;
/*
* There are lots of broken chipsets in the world. Some of them will
* violate the PCI spec when we issue byte sized memory writes to our
* controller. I/O mapped register access, if allowed by the given
* platform, will work in almost all cases.
*/
uint32_t aic79xx_allow_memio = ~0;
/*
* aic79xx_detect() has been run, so register all device arrivals
* immediately with the system rather than deferring to the sorted
* attachment performed by aic79xx_detect().
*/
int aic79xx_detect_complete;
/*
* So that we can set how long each device is given as a selection timeout.
* The table of values goes like this:
* 0 - 256ms
* 1 - 128ms
* 2 - 64ms
* 3 - 32ms
* We default to 256ms because some older devices need a longer time
* to respond to initial selection.
*/
static uint32_t aic79xx_seltime;
/*
* Certain devices do not perform any aging on commands. Should the
* device be saturated by commands in one portion of the disk, it is
* possible for transactions on far away sectors to never be serviced.
* To handle these devices, we can periodically send an ordered tag to
* force all outstanding transactions to be serviced prior to a new
* transaction.
*/
uint32_t aic79xx_periodic_otag;
/*
* Module information and settable options.
*/
static char *aic79xx = NULL;
MODULE_AUTHOR("Maintainer: Justin T. Gibbs <gibbs@scsiguy.com>");
MODULE_DESCRIPTION("Adaptec Aic790X U320 SCSI Host Bus Adapter driver");
MODULE_LICENSE("Dual BSD/GPL");
MODULE_VERSION(AIC79XX_DRIVER_VERSION);
module_param(aic79xx, charp, 0);
MODULE_PARM_DESC(aic79xx,
"period delimited, options string.\n"
" verbose Enable verbose/diagnostic logging\n"
" allow_memio Allow device registers to be memory mapped\n"
" debug Bitmask of debug values to enable\n"
" no_reset Supress initial bus resets\n"
" extended Enable extended geometry on all controllers\n"
" periodic_otag Send an ordered tagged transaction\n"
" periodically to prevent tag starvation.\n"
" This may be required by some older disk\n"
" or drives/RAID arrays.\n"
" reverse_scan Sort PCI devices highest Bus/Slot to lowest\n"
" tag_info:<tag_str> Set per-target tag depth\n"
" global_tag_depth:<int> Global tag depth for all targets on all buses\n"
" rd_strm:<rd_strm_masks> Set per-target read streaming setting.\n"
" dv:<dv_settings> Set per-controller Domain Validation Setting.\n"
" slewrate:<slewrate_list>Set the signal slew rate (0-15).\n"
" precomp:<pcomp_list> Set the signal precompensation (0-7).\n"
" amplitude:<int> Set the signal amplitude (0-7).\n"
" seltime:<int> Selection Timeout:\n"
" (0/256ms,1/128ms,2/64ms,3/32ms)\n"
"\n"
" Sample /etc/modprobe.conf line:\n"
" Enable verbose logging\n"
" Set tag depth on Controller 2/Target 2 to 10 tags\n"
" Shorten the selection timeout to 128ms\n"
"\n"
" options aic79xx 'aic79xx=verbose.tag_info:{{}.{}.{..10}}.seltime:1'\n"
"\n"
" Sample /etc/modprobe.conf line:\n"
" Change Read Streaming for Controller's 2 and 3\n"
"\n"
" options aic79xx 'aic79xx=rd_strm:{..0xFFF0.0xC0F0}'");
static void ahd_linux_handle_scsi_status(struct ahd_softc *,
struct ahd_linux_device *,
struct scb *);
static void ahd_linux_queue_cmd_complete(struct ahd_softc *ahd,
Scsi_Cmnd *cmd);
static void ahd_linux_filter_inquiry(struct ahd_softc *ahd,
struct ahd_devinfo *devinfo);
static void ahd_linux_dev_timed_unfreeze(u_long arg);
static void ahd_linux_sem_timeout(u_long arg);
static void ahd_linux_initialize_scsi_bus(struct ahd_softc *ahd);
static void ahd_linux_size_nseg(void);
static void ahd_linux_thread_run_complete_queue(struct ahd_softc *ahd);
static void ahd_linux_start_dv(struct ahd_softc *ahd);
static void ahd_linux_dv_timeout(struct scsi_cmnd *cmd);
static int ahd_linux_dv_thread(void *data);
static void ahd_linux_kill_dv_thread(struct ahd_softc *ahd);
static void ahd_linux_dv_target(struct ahd_softc *ahd, u_int target);
static void ahd_linux_dv_transition(struct ahd_softc *ahd,
struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo,
struct ahd_linux_target *targ);
static void ahd_linux_dv_fill_cmd(struct ahd_softc *ahd,
struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo);
static void ahd_linux_dv_inq(struct ahd_softc *ahd,
struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo,
struct ahd_linux_target *targ,
u_int request_length);
static void ahd_linux_dv_tur(struct ahd_softc *ahd,
struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo);
static void ahd_linux_dv_rebd(struct ahd_softc *ahd,
struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo,
struct ahd_linux_target *targ);
static void ahd_linux_dv_web(struct ahd_softc *ahd,
struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo,
struct ahd_linux_target *targ);
static void ahd_linux_dv_reb(struct ahd_softc *ahd,
struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo,
struct ahd_linux_target *targ);
static void ahd_linux_dv_su(struct ahd_softc *ahd,
struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo,
struct ahd_linux_target *targ);
static int ahd_linux_fallback(struct ahd_softc *ahd,
struct ahd_devinfo *devinfo);
static __inline int ahd_linux_dv_fallback(struct ahd_softc *ahd,
struct ahd_devinfo *devinfo);
static void ahd_linux_dv_complete(Scsi_Cmnd *cmd);
static void ahd_linux_generate_dv_pattern(struct ahd_linux_target *targ);
static u_int ahd_linux_user_tagdepth(struct ahd_softc *ahd,
struct ahd_devinfo *devinfo);
static u_int ahd_linux_user_dv_setting(struct ahd_softc *ahd);
static void ahd_linux_setup_user_rd_strm_settings(struct ahd_softc *ahd);
static void ahd_linux_device_queue_depth(struct ahd_softc *ahd,
struct ahd_linux_device *dev);
static struct ahd_linux_target* ahd_linux_alloc_target(struct ahd_softc*,
u_int, u_int);
static void ahd_linux_free_target(struct ahd_softc*,
struct ahd_linux_target*);
static struct ahd_linux_device* ahd_linux_alloc_device(struct ahd_softc*,
struct ahd_linux_target*,
u_int);
static void ahd_linux_free_device(struct ahd_softc*,
struct ahd_linux_device*);
static void ahd_linux_run_device_queue(struct ahd_softc*,
struct ahd_linux_device*);
static void ahd_linux_setup_tag_info_global(char *p);
static aic_option_callback_t ahd_linux_setup_tag_info;
static aic_option_callback_t ahd_linux_setup_rd_strm_info;
static aic_option_callback_t ahd_linux_setup_dv;
static aic_option_callback_t ahd_linux_setup_iocell_info;
static int ahd_linux_next_unit(void);
static void ahd_runq_tasklet(unsigned long data);
static int aic79xx_setup(char *c);
/****************************** Inlines ***************************************/
static __inline void ahd_schedule_completeq(struct ahd_softc *ahd);
static __inline void ahd_schedule_runq(struct ahd_softc *ahd);
static __inline void ahd_setup_runq_tasklet(struct ahd_softc *ahd);
static __inline void ahd_teardown_runq_tasklet(struct ahd_softc *ahd);
static __inline struct ahd_linux_device*
ahd_linux_get_device(struct ahd_softc *ahd, u_int channel,
u_int target, u_int lun, int alloc);
static struct ahd_cmd *ahd_linux_run_complete_queue(struct ahd_softc *ahd);
static __inline void ahd_linux_check_device_queue(struct ahd_softc *ahd,
struct ahd_linux_device *dev);
static __inline struct ahd_linux_device *
ahd_linux_next_device_to_run(struct ahd_softc *ahd);
static __inline void ahd_linux_run_device_queues(struct ahd_softc *ahd);
static __inline void ahd_linux_unmap_scb(struct ahd_softc*, struct scb*);
static __inline void
ahd_schedule_completeq(struct ahd_softc *ahd)
{
if ((ahd->platform_data->flags & AHD_RUN_CMPLT_Q_TIMER) == 0) {
ahd->platform_data->flags |= AHD_RUN_CMPLT_Q_TIMER;
ahd->platform_data->completeq_timer.expires = jiffies;
add_timer(&ahd->platform_data->completeq_timer);
}
}
/*
* Must be called with our lock held.
*/
static __inline void
ahd_schedule_runq(struct ahd_softc *ahd)
{
tasklet_schedule(&ahd->platform_data->runq_tasklet);
}
static __inline
void ahd_setup_runq_tasklet(struct ahd_softc *ahd)
{
tasklet_init(&ahd->platform_data->runq_tasklet, ahd_runq_tasklet,
(unsigned long)ahd);
}
static __inline void
ahd_teardown_runq_tasklet(struct ahd_softc *ahd)
{
tasklet_kill(&ahd->platform_data->runq_tasklet);
}
static __inline struct ahd_linux_device*
ahd_linux_get_device(struct ahd_softc *ahd, u_int channel, u_int target,
u_int lun, int alloc)
{
struct ahd_linux_target *targ;
struct ahd_linux_device *dev;
u_int target_offset;
target_offset = target;
if (channel != 0)
target_offset += 8;
targ = ahd->platform_data->targets[target_offset];
if (targ == NULL) {
if (alloc != 0) {
targ = ahd_linux_alloc_target(ahd, channel, target);
if (targ == NULL)
return (NULL);
} else
return (NULL);
}
dev = targ->devices[lun];
if (dev == NULL && alloc != 0)
dev = ahd_linux_alloc_device(ahd, targ, lun);
return (dev);
}
#define AHD_LINUX_MAX_RETURNED_ERRORS 4
static struct ahd_cmd *
ahd_linux_run_complete_queue(struct ahd_softc *ahd)
{
struct ahd_cmd *acmd;
u_long done_flags;
int with_errors;
with_errors = 0;
ahd_done_lock(ahd, &done_flags);
while ((acmd = TAILQ_FIRST(&ahd->platform_data->completeq)) != NULL) {
Scsi_Cmnd *cmd;
if (with_errors > AHD_LINUX_MAX_RETURNED_ERRORS) {
/*
* Linux uses stack recursion to requeue
* commands that need to be retried. Avoid
* blowing out the stack by "spoon feeding"
* commands that completed with error back
* the operating system in case they are going
* to be retried. "ick"
*/
ahd_schedule_completeq(ahd);
break;
}
TAILQ_REMOVE(&ahd->platform_data->completeq,
acmd, acmd_links.tqe);
cmd = &acmd_scsi_cmd(acmd);
cmd->host_scribble = NULL;
if (ahd_cmd_get_transaction_status(cmd) != DID_OK
|| (cmd->result & 0xFF) != SCSI_STATUS_OK)
with_errors++;
cmd->scsi_done(cmd);
}
ahd_done_unlock(ahd, &done_flags);
return (acmd);
}
static __inline void
ahd_linux_check_device_queue(struct ahd_softc *ahd,
struct ahd_linux_device *dev)
{
if ((dev->flags & AHD_DEV_FREEZE_TIL_EMPTY) != 0
&& dev->active == 0) {
dev->flags &= ~AHD_DEV_FREEZE_TIL_EMPTY;
dev->qfrozen--;
}
if (TAILQ_FIRST(&dev->busyq) == NULL
|| dev->openings == 0 || dev->qfrozen != 0)
return;
ahd_linux_run_device_queue(ahd, dev);
}
static __inline struct ahd_linux_device *
ahd_linux_next_device_to_run(struct ahd_softc *ahd)
{
if ((ahd->flags & AHD_RESOURCE_SHORTAGE) != 0
|| (ahd->platform_data->qfrozen != 0
&& AHD_DV_SIMQ_FROZEN(ahd) == 0))
return (NULL);
return (TAILQ_FIRST(&ahd->platform_data->device_runq));
}
static __inline void
ahd_linux_run_device_queues(struct ahd_softc *ahd)
{
struct ahd_linux_device *dev;
while ((dev = ahd_linux_next_device_to_run(ahd)) != NULL) {
TAILQ_REMOVE(&ahd->platform_data->device_runq, dev, links);
dev->flags &= ~AHD_DEV_ON_RUN_LIST;
ahd_linux_check_device_queue(ahd, dev);
}
}
static __inline void
ahd_linux_unmap_scb(struct ahd_softc *ahd, struct scb *scb)
{
Scsi_Cmnd *cmd;
int direction;
cmd = scb->io_ctx;
direction = cmd->sc_data_direction;
ahd_sync_sglist(ahd, scb, BUS_DMASYNC_POSTWRITE);
if (cmd->use_sg != 0) {
struct scatterlist *sg;
sg = (struct scatterlist *)cmd->request_buffer;
pci_unmap_sg(ahd->dev_softc, sg, cmd->use_sg, direction);
} else if (cmd->request_bufflen != 0) {
pci_unmap_single(ahd->dev_softc,
scb->platform_data->buf_busaddr,
cmd->request_bufflen, direction);
}
}
/******************************** Macros **************************************/
#define BUILD_SCSIID(ahd, cmd) \
((((cmd)->device->id << TID_SHIFT) & TID) | (ahd)->our_id)
/************************ Host template entry points *************************/
static int ahd_linux_detect(Scsi_Host_Template *);
static const char *ahd_linux_info(struct Scsi_Host *);
static int ahd_linux_queue(Scsi_Cmnd *, void (*)(Scsi_Cmnd *));
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
static int ahd_linux_slave_alloc(Scsi_Device *);
static int ahd_linux_slave_configure(Scsi_Device *);
static void ahd_linux_slave_destroy(Scsi_Device *);
#if defined(__i386__)
static int ahd_linux_biosparam(struct scsi_device*,
struct block_device*, sector_t, int[]);
#endif
#else
static int ahd_linux_release(struct Scsi_Host *);
static void ahd_linux_select_queue_depth(struct Scsi_Host *host,
Scsi_Device *scsi_devs);
#if defined(__i386__)
static int ahd_linux_biosparam(Disk *, kdev_t, int[]);
#endif
#endif
static int ahd_linux_bus_reset(Scsi_Cmnd *);
static int ahd_linux_dev_reset(Scsi_Cmnd *);
static int ahd_linux_abort(Scsi_Cmnd *);
/*
* Calculate a safe value for AHD_NSEG (as expressed through ahd_linux_nseg).
*
* In pre-2.5.X...
* The midlayer allocates an S/G array dynamically when a command is issued
* using SCSI malloc. This array, which is in an OS dependent format that
* must later be copied to our private S/G list, is sized to house just the
* number of segments needed for the current transfer. Since the code that
* sizes the SCSI malloc pool does not take into consideration fragmentation
* of the pool, executing transactions numbering just a fraction of our
* concurrent transaction limit with SG list lengths aproaching AHC_NSEG will
* quickly depleat the SCSI malloc pool of usable space. Unfortunately, the
* mid-layer does not properly handle this scsi malloc failures for the S/G
* array and the result can be a lockup of the I/O subsystem. We try to size
* our S/G list so that it satisfies our drivers allocation requirements in
* addition to avoiding fragmentation of the SCSI malloc pool.
*/
static void
ahd_linux_size_nseg(void)
{
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
u_int cur_size;
u_int best_size;
/*
* The SCSI allocator rounds to the nearest 512 bytes
* an cannot allocate across a page boundary. Our algorithm
* is to start at 1K of scsi malloc space per-command and
* loop through all factors of the PAGE_SIZE and pick the best.
*/
best_size = 0;
for (cur_size = 1024; cur_size <= PAGE_SIZE; cur_size *= 2) {
u_int nseg;
nseg = cur_size / sizeof(struct scatterlist);
if (nseg < AHD_LINUX_MIN_NSEG)
continue;
if (best_size == 0) {
best_size = cur_size;
ahd_linux_nseg = nseg;
} else {
u_int best_rem;
u_int cur_rem;
/*
* Compare the traits of the current "best_size"
* with the current size to determine if the
* current size is a better size.
*/
best_rem = best_size % sizeof(struct scatterlist);
cur_rem = cur_size % sizeof(struct scatterlist);
if (cur_rem < best_rem) {
best_size = cur_size;
ahd_linux_nseg = nseg;
}
}
}
#endif
}
/*
* Try to detect an Adaptec 79XX controller.
*/
static int
ahd_linux_detect(Scsi_Host_Template *template)
{
struct ahd_softc *ahd;
int found;
int error = 0;
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
/*
* It is a bug that the upper layer takes
* this lock just prior to calling us.
*/
spin_unlock_irq(&io_request_lock);
#endif
/*
* Sanity checking of Linux SCSI data structures so
* that some of our hacks^H^H^H^H^Hassumptions aren't
* violated.
*/
if (offsetof(struct ahd_cmd_internal, end)
> offsetof(struct scsi_cmnd, host_scribble)) {
printf("ahd_linux_detect: SCSI data structures changed.\n");
printf("ahd_linux_detect: Unable to attach\n");
return (0);
}
/*
* Determine an appropriate size for our Scatter Gatther lists.
*/
ahd_linux_size_nseg();
#ifdef MODULE
/*
* If we've been passed any parameters, process them now.
*/
if (aic79xx)
aic79xx_setup(aic79xx);
#endif
template->proc_name = "aic79xx";
/*
* Initialize our softc list lock prior to
* probing for any adapters.
*/
ahd_list_lockinit();
#ifdef CONFIG_PCI
error = ahd_linux_pci_init();
if (error)
return error;
#endif
/*
* Register with the SCSI layer all
* controllers we've found.
*/
found = 0;
TAILQ_FOREACH(ahd, &ahd_tailq, links) {
if (ahd_linux_register_host(ahd, template) == 0)
found++;
}
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
spin_lock_irq(&io_request_lock);
#endif
aic79xx_detect_complete++;
return 0;
}
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
/*
* Free the passed in Scsi_Host memory structures prior to unloading the
* module.
*/
static int
ahd_linux_release(struct Scsi_Host * host)
{
struct ahd_softc *ahd;
u_long l;
ahd_list_lock(&l);
if (host != NULL) {
/*
* We should be able to just perform
* the free directly, but check our
* list for extra sanity.
*/
ahd = ahd_find_softc(*(struct ahd_softc **)host->hostdata);
if (ahd != NULL) {
u_long s;
ahd_lock(ahd, &s);
ahd_intr_enable(ahd, FALSE);
ahd_unlock(ahd, &s);
ahd_free(ahd);
}
}
ahd_list_unlock(&l);
return (0);
}
#endif
/*
* Return a string describing the driver.
*/
static const char *
ahd_linux_info(struct Scsi_Host *host)
{
static char buffer[512];
char ahd_info[256];
char *bp;
struct ahd_softc *ahd;
bp = &buffer[0];
ahd = *(struct ahd_softc **)host->hostdata;
memset(bp, 0, sizeof(buffer));
strcpy(bp, "Adaptec AIC79XX PCI-X SCSI HBA DRIVER, Rev ");
strcat(bp, AIC79XX_DRIVER_VERSION);
strcat(bp, "\n");
strcat(bp, " <");
strcat(bp, ahd->description);
strcat(bp, ">\n");
strcat(bp, " ");
ahd_controller_info(ahd, ahd_info);
strcat(bp, ahd_info);
strcat(bp, "\n");
return (bp);
}
/*
* Queue an SCB to the controller.
*/
static int
ahd_linux_queue(Scsi_Cmnd * cmd, void (*scsi_done) (Scsi_Cmnd *))
{
struct ahd_softc *ahd;
struct ahd_linux_device *dev;
u_long flags;
ahd = *(struct ahd_softc **)cmd->device->host->hostdata;
/*
* Save the callback on completion function.
*/
cmd->scsi_done = scsi_done;
ahd_midlayer_entrypoint_lock(ahd, &flags);
/*
* Close the race of a command that was in the process of
* being queued to us just as our simq was frozen. Let
* DV commands through so long as we are only frozen to
* perform DV.
*/
if (ahd->platform_data->qfrozen != 0
&& AHD_DV_CMD(cmd) == 0) {
ahd_cmd_set_transaction_status(cmd, CAM_REQUEUE_REQ);
ahd_linux_queue_cmd_complete(ahd, cmd);
ahd_schedule_completeq(ahd);
ahd_midlayer_entrypoint_unlock(ahd, &flags);
return (0);
}
dev = ahd_linux_get_device(ahd, cmd->device->channel,
cmd->device->id, cmd->device->lun,
/*alloc*/TRUE);
if (dev == NULL) {
ahd_cmd_set_transaction_status(cmd, CAM_RESRC_UNAVAIL);
ahd_linux_queue_cmd_complete(ahd, cmd);
ahd_schedule_completeq(ahd);
ahd_midlayer_entrypoint_unlock(ahd, &flags);
printf("%s: aic79xx_linux_queue - Unable to allocate device!\n",
ahd_name(ahd));
return (0);
}
if (cmd->cmd_len > MAX_CDB_LEN)
return (-EINVAL);
cmd->result = CAM_REQ_INPROG << 16;
TAILQ_INSERT_TAIL(&dev->busyq, (struct ahd_cmd *)cmd, acmd_links.tqe);
if ((dev->flags & AHD_DEV_ON_RUN_LIST) == 0) {
TAILQ_INSERT_TAIL(&ahd->platform_data->device_runq, dev, links);
dev->flags |= AHD_DEV_ON_RUN_LIST;
ahd_linux_run_device_queues(ahd);
}
ahd_midlayer_entrypoint_unlock(ahd, &flags);
return (0);
}
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
static int
ahd_linux_slave_alloc(Scsi_Device *device)
{
struct ahd_softc *ahd;
ahd = *((struct ahd_softc **)device->host->hostdata);
if (bootverbose)
printf("%s: Slave Alloc %d\n", ahd_name(ahd), device->id);
return (0);
}
static int
ahd_linux_slave_configure(Scsi_Device *device)
{
struct ahd_softc *ahd;
struct ahd_linux_device *dev;
u_long flags;
ahd = *((struct ahd_softc **)device->host->hostdata);
if (bootverbose)
printf("%s: Slave Configure %d\n", ahd_name(ahd), device->id);
ahd_midlayer_entrypoint_lock(ahd, &flags);
/*
* Since Linux has attached to the device, configure
* it so we don't free and allocate the device
* structure on every command.
*/
dev = ahd_linux_get_device(ahd, device->channel,
device->id, device->lun,
/*alloc*/TRUE);
if (dev != NULL) {
dev->flags &= ~AHD_DEV_UNCONFIGURED;
dev->flags |= AHD_DEV_SLAVE_CONFIGURED;
dev->scsi_device = device;
ahd_linux_device_queue_depth(ahd, dev);
}
ahd_midlayer_entrypoint_unlock(ahd, &flags);
return (0);
}
static void
ahd_linux_slave_destroy(Scsi_Device *device)
{
struct ahd_softc *ahd;
struct ahd_linux_device *dev;
u_long flags;
ahd = *((struct ahd_softc **)device->host->hostdata);
if (bootverbose)
printf("%s: Slave Destroy %d\n", ahd_name(ahd), device->id);
ahd_midlayer_entrypoint_lock(ahd, &flags);
dev = ahd_linux_get_device(ahd, device->channel,
device->id, device->lun,
/*alloc*/FALSE);
/*
* Filter out "silly" deletions of real devices by only
* deleting devices that have had slave_configure()
* called on them. All other devices that have not
* been configured will automatically be deleted by
* the refcounting process.
*/
if (dev != NULL
&& (dev->flags & AHD_DEV_SLAVE_CONFIGURED) != 0) {
dev->flags |= AHD_DEV_UNCONFIGURED;
if (TAILQ_EMPTY(&dev->busyq)
&& dev->active == 0
&& (dev->flags & AHD_DEV_TIMER_ACTIVE) == 0)
ahd_linux_free_device(ahd, dev);
}
ahd_midlayer_entrypoint_unlock(ahd, &flags);
}
#else
/*
* Sets the queue depth for each SCSI device hanging
* off the input host adapter.
*/
static void
ahd_linux_select_queue_depth(struct Scsi_Host * host,
Scsi_Device * scsi_devs)
{
Scsi_Device *device;
Scsi_Device *ldev;
struct ahd_softc *ahd;
u_long flags;
ahd = *((struct ahd_softc **)host->hostdata);
ahd_lock(ahd, &flags);
for (device = scsi_devs; device != NULL; device = device->next) {
/*
* Watch out for duplicate devices. This works around
* some quirks in how the SCSI scanning code does its
* device management.
*/
for (ldev = scsi_devs; ldev != device; ldev = ldev->next) {
if (ldev->host == device->host
&& ldev->channel == device->channel
&& ldev->id == device->id
&& ldev->lun == device->lun)
break;
}
/* Skip duplicate. */
if (ldev != device)
continue;
if (device->host == host) {
struct ahd_linux_device *dev;
/*
* Since Linux has attached to the device, configure
* it so we don't free and allocate the device
* structure on every command.
*/
dev = ahd_linux_get_device(ahd, device->channel,
device->id, device->lun,
/*alloc*/TRUE);
if (dev != NULL) {
dev->flags &= ~AHD_DEV_UNCONFIGURED;
dev->scsi_device = device;
ahd_linux_device_queue_depth(ahd, dev);
device->queue_depth = dev->openings
+ dev->active;
if ((dev->flags & (AHD_DEV_Q_BASIC
| AHD_DEV_Q_TAGGED)) == 0) {
/*
* We allow the OS to queue 2 untagged
* transactions to us at any time even
* though we can only execute them
* serially on the controller/device.
* This should remove some latency.
*/
device->queue_depth = 2;
}
}
}
}
ahd_unlock(ahd, &flags);
}
#endif
#if defined(__i386__)
/*
* Return the disk geometry for the given SCSI device.
*/
static int
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
ahd_linux_biosparam(struct scsi_device *sdev, struct block_device *bdev,
sector_t capacity, int geom[])
{
uint8_t *bh;
#else
ahd_linux_biosparam(Disk *disk, kdev_t dev, int geom[])
{
struct scsi_device *sdev = disk->device;
u_long capacity = disk->capacity;
struct buffer_head *bh;
#endif
int heads;
int sectors;
int cylinders;
int ret;
int extended;
struct ahd_softc *ahd;
ahd = *((struct ahd_softc **)sdev->host->hostdata);
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
bh = scsi_bios_ptable(bdev);
#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,17)
bh = bread(MKDEV(MAJOR(dev), MINOR(dev) & ~0xf), 0, block_size(dev));
#else
bh = bread(MKDEV(MAJOR(dev), MINOR(dev) & ~0xf), 0, 1024);
#endif
if (bh) {
ret = scsi_partsize(bh, capacity,
&geom[2], &geom[0], &geom[1]);
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
kfree(bh);
#else
brelse(bh);
#endif
if (ret != -1)
return (ret);
}
heads = 64;
sectors = 32;
cylinders = aic_sector_div(capacity, heads, sectors);
if (aic79xx_extended != 0)
extended = 1;
else
extended = (ahd->flags & AHD_EXTENDED_TRANS_A) != 0;
if (extended && cylinders >= 1024) {
heads = 255;
sectors = 63;
cylinders = aic_sector_div(capacity, heads, sectors);
}
geom[0] = heads;
geom[1] = sectors;
geom[2] = cylinders;
return (0);
}
#endif
/*
* Abort the current SCSI command(s).
*/
static int
ahd_linux_abort(Scsi_Cmnd *cmd)
{
struct ahd_softc *ahd;
struct ahd_cmd *acmd;
struct ahd_cmd *list_acmd;
struct ahd_linux_device *dev;
struct scb *pending_scb;
u_long s;
u_int saved_scbptr;
u_int active_scbptr;
u_int last_phase;
u_int cdb_byte;
int retval;
int was_paused;
int paused;
int wait;
int disconnected;
ahd_mode_state saved_modes;
pending_scb = NULL;
paused = FALSE;
wait = FALSE;
ahd = *(struct ahd_softc **)cmd->device->host->hostdata;
acmd = (struct ahd_cmd *)cmd;
printf("%s:%d:%d:%d: Attempting to abort cmd %p:",
ahd_name(ahd), cmd->device->channel, cmd->device->id,
cmd->device->lun, cmd);
for (cdb_byte = 0; cdb_byte < cmd->cmd_len; cdb_byte++)
printf(" 0x%x", cmd->cmnd[cdb_byte]);
printf("\n");
/*
* In all versions of Linux, we have to work around
* a major flaw in how the mid-layer is locked down
* if we are to sleep successfully in our error handler
* while allowing our interrupt handler to run. Since
* the midlayer acquires either the io_request_lock or
* our lock prior to calling us, we must use the
* spin_unlock_irq() method for unlocking our lock.
* This will force interrupts to be enabled on the
* current CPU. Since the EH thread should not have
* been running with CPU interrupts disabled other than
* by acquiring either the io_request_lock or our own
* lock, this *should* be safe.
*/
ahd_midlayer_entrypoint_lock(ahd, &s);
/*
* First determine if we currently own this command.
* Start by searching the device queue. If not found
* there, check the pending_scb list. If not found
* at all, and the system wanted us to just abort the
* command, return success.
*/
dev = ahd_linux_get_device(ahd, cmd->device->channel,
cmd->device->id, cmd->device->lun,
/*alloc*/FALSE);
if (dev == NULL) {
/*
* No target device for this command exists,
* so we must not still own the command.
*/
printf("%s:%d:%d:%d: Is not an active device\n",
ahd_name(ahd), cmd->device->channel, cmd->device->id,
cmd->device->lun);
retval = SUCCESS;
goto no_cmd;
}
TAILQ_FOREACH(list_acmd, &dev->busyq, acmd_links.tqe) {
if (list_acmd == acmd)
break;
}
if (list_acmd != NULL) {
printf("%s:%d:%d:%d: Command found on device queue\n",
ahd_name(ahd), cmd->device->channel, cmd->device->id,
cmd->device->lun);
TAILQ_REMOVE(&dev->busyq, list_acmd, acmd_links.tqe);
cmd->result = DID_ABORT << 16;
ahd_linux_queue_cmd_complete(ahd, cmd);
retval = SUCCESS;
goto done;
}
/*
* See if we can find a matching cmd in the pending list.
*/
LIST_FOREACH(pending_scb, &ahd->pending_scbs, pending_links) {
if (pending_scb->io_ctx == cmd)
break;
}
if (pending_scb == NULL) {
printf("%s:%d:%d:%d: Command not found\n",
ahd_name(ahd), cmd->device->channel, cmd->device->id,
cmd->device->lun);
goto no_cmd;
}
if ((pending_scb->flags & SCB_RECOVERY_SCB) != 0) {
/*
* We can't queue two recovery actions using the same SCB
*/
retval = FAILED;
goto done;
}
/*
* Ensure that the card doesn't do anything
* behind our back. Also make sure that we
* didn't "just" miss an interrupt that would
* affect this cmd.
*/
was_paused = ahd_is_paused(ahd);
ahd_pause_and_flushwork(ahd);
paused = TRUE;
if ((pending_scb->flags & SCB_ACTIVE) == 0) {
printf("%s:%d:%d:%d: Command already completed\n",
ahd_name(ahd), cmd->device->channel, cmd->device->id,
cmd->device->lun);
goto no_cmd;
}
printf("%s: At time of recovery, card was %spaused\n",
ahd_name(ahd), was_paused ? "" : "not ");
ahd_dump_card_state(ahd);
disconnected = TRUE;
if (ahd_search_qinfifo(ahd, cmd->device->id, cmd->device->channel + 'A',
cmd->device->lun, SCB_GET_TAG(pending_scb),
ROLE_INITIATOR, CAM_REQ_ABORTED,
SEARCH_COMPLETE) > 0) {
printf("%s:%d:%d:%d: Cmd aborted from QINFIFO\n",
ahd_name(ahd), cmd->device->channel, cmd->device->id,
cmd->device->lun);
retval = SUCCESS;
goto done;
}
saved_modes = ahd_save_modes(ahd);
ahd_set_modes(ahd, AHD_MODE_SCSI, AHD_MODE_SCSI);
last_phase = ahd_inb(ahd, LASTPHASE);
saved_scbptr = ahd_get_scbptr(ahd);
active_scbptr = saved_scbptr;
if (disconnected && (ahd_inb(ahd, SEQ_FLAGS) & NOT_IDENTIFIED) == 0) {
struct scb *bus_scb;
bus_scb = ahd_lookup_scb(ahd, active_scbptr);
if (bus_scb == pending_scb)
disconnected = FALSE;
}
/*
* At this point, pending_scb is the scb associated with the
* passed in command. That command is currently active on the
* bus or is in the disconnected state.
*/
if (last_phase != P_BUSFREE
&& SCB_GET_TAG(pending_scb) == active_scbptr) {
/*
* We're active on the bus, so assert ATN
* and hope that the target responds.
*/
pending_scb = ahd_lookup_scb(ahd, active_scbptr);
pending_scb->flags |= SCB_RECOVERY_SCB|SCB_ABORT;
ahd_outb(ahd, MSG_OUT, HOST_MSG);
ahd_outb(ahd, SCSISIGO, last_phase|ATNO);
printf("%s:%d:%d:%d: Device is active, asserting ATN\n",
ahd_name(ahd), cmd->device->channel,
cmd->device->id, cmd->device->lun);
wait = TRUE;
} else if (disconnected) {
/*
* Actually re-queue this SCB in an attempt
* to select the device before it reconnects.
*/
pending_scb->flags |= SCB_RECOVERY_SCB|SCB_ABORT;
ahd_set_scbptr(ahd, SCB_GET_TAG(pending_scb));
pending_scb->hscb->cdb_len = 0;
pending_scb->hscb->task_attribute = 0;
pending_scb->hscb->task_management = SIU_TASKMGMT_ABORT_TASK;
if ((pending_scb->flags & SCB_PACKETIZED) != 0) {
/*
* Mark the SCB has having an outstanding
* task management function. Should the command
* complete normally before the task management
* function can be sent, the host will be notified
* to abort our requeued SCB.
*/
ahd_outb(ahd, SCB_TASK_MANAGEMENT,
pending_scb->hscb->task_management);
} else {
/*
* If non-packetized, set the MK_MESSAGE control
* bit indicating that we desire to send a message.
* We also set the disconnected flag since there is
* no guarantee that our SCB control byte matches
* the version on the card. We don't want the
* sequencer to abort the command thinking an
* unsolicited reselection occurred.
*/
pending_scb->hscb->control |= MK_MESSAGE|DISCONNECTED;
/*
* The sequencer will never re-reference the
* in-core SCB. To make sure we are notified
* during reslection, set the MK_MESSAGE flag in
* the card's copy of the SCB.
*/
ahd_outb(ahd, SCB_CONTROL,
ahd_inb(ahd, SCB_CONTROL)|MK_MESSAGE);
}
/*
* Clear out any entries in the QINFIFO first
* so we are the next SCB for this target
* to run.
*/
ahd_search_qinfifo(ahd, cmd->device->id,
cmd->device->channel + 'A', cmd->device->lun,
SCB_LIST_NULL, ROLE_INITIATOR,
CAM_REQUEUE_REQ, SEARCH_COMPLETE);
ahd_qinfifo_requeue_tail(ahd, pending_scb);
ahd_set_scbptr(ahd, saved_scbptr);
ahd_print_path(ahd, pending_scb);
printf("Device is disconnected, re-queuing SCB\n");
wait = TRUE;
} else {
printf("%s:%d:%d:%d: Unable to deliver message\n",
ahd_name(ahd), cmd->device->channel,
cmd->device->id, cmd->device->lun);
retval = FAILED;
goto done;
}
no_cmd:
/*
* Our assumption is that if we don't have the command, no
* recovery action was required, so we return success. Again,
* the semantics of the mid-layer recovery engine are not
* well defined, so this may change in time.
*/
retval = SUCCESS;
done:
if (paused)
ahd_unpause(ahd);
if (wait) {
struct timer_list timer;
int ret;
pending_scb->platform_data->flags |= AHD_SCB_UP_EH_SEM;
spin_unlock_irq(&ahd->platform_data->spin_lock);
init_timer(&timer);
timer.data = (u_long)pending_scb;
timer.expires = jiffies + (5 * HZ);
timer.function = ahd_linux_sem_timeout;
add_timer(&timer);
printf("Recovery code sleeping\n");
down(&ahd->platform_data->eh_sem);
printf("Recovery code awake\n");
ret = del_timer_sync(&timer);
if (ret == 0) {
printf("Timer Expired\n");
retval = FAILED;
}
spin_lock_irq(&ahd->platform_data->spin_lock);
}
ahd_schedule_runq(ahd);
ahd_linux_run_complete_queue(ahd);
ahd_midlayer_entrypoint_unlock(ahd, &s);
return (retval);
}
static void
ahd_linux_dev_reset_complete(Scsi_Cmnd *cmd)
{
free(cmd, M_DEVBUF);
}
/*
* Attempt to send a target reset message to the device that timed out.
*/
static int
ahd_linux_dev_reset(Scsi_Cmnd *cmd)
{
struct ahd_softc *ahd;
struct scsi_cmnd *recovery_cmd;
struct ahd_linux_device *dev;
struct ahd_initiator_tinfo *tinfo;
struct ahd_tmode_tstate *tstate;
struct scb *scb;
struct hardware_scb *hscb;
u_long s;
struct timer_list timer;
int retval;
ahd = *(struct ahd_softc **)cmd->device->host->hostdata;
recovery_cmd = malloc(sizeof(struct scsi_cmnd), M_DEVBUF, M_WAITOK);
if (!recovery_cmd)
return (FAILED);
memset(recovery_cmd, 0, sizeof(struct scsi_cmnd));
recovery_cmd->device = cmd->device;
recovery_cmd->scsi_done = ahd_linux_dev_reset_complete;
#if AHD_DEBUG
if ((ahd_debug & AHD_SHOW_RECOVERY) != 0)
printf("%s:%d:%d:%d: Device reset called for cmd %p\n",
ahd_name(ahd), cmd->device->channel, cmd->device->id,
cmd->device->lun, cmd);
#endif
ahd_midlayer_entrypoint_lock(ahd, &s);
dev = ahd_linux_get_device(ahd, cmd->device->channel, cmd->device->id,
cmd->device->lun, /*alloc*/FALSE);
if (dev == NULL) {
ahd_midlayer_entrypoint_unlock(ahd, &s);
kfree(recovery_cmd);
return (FAILED);
}
if ((scb = ahd_get_scb(ahd, AHD_NEVER_COL_IDX)) == NULL) {
ahd_midlayer_entrypoint_unlock(ahd, &s);
kfree(recovery_cmd);
return (FAILED);
}
tinfo = ahd_fetch_transinfo(ahd, 'A', ahd->our_id,
cmd->device->id, &tstate);
recovery_cmd->result = CAM_REQ_INPROG << 16;
recovery_cmd->host_scribble = (char *)scb;
scb->io_ctx = recovery_cmd;
scb->platform_data->dev = dev;
scb->sg_count = 0;
ahd_set_residual(scb, 0);
ahd_set_sense_residual(scb, 0);
hscb = scb->hscb;
hscb->control = 0;
hscb->scsiid = BUILD_SCSIID(ahd, cmd);
hscb->lun = cmd->device->lun;
hscb->cdb_len = 0;
hscb->task_management = SIU_TASKMGMT_LUN_RESET;
scb->flags |= SCB_DEVICE_RESET|SCB_RECOVERY_SCB|SCB_ACTIVE;
if ((tinfo->curr.ppr_options & MSG_EXT_PPR_IU_REQ) != 0) {
scb->flags |= SCB_PACKETIZED;
} else {
hscb->control |= MK_MESSAGE;
}
dev->openings--;
dev->active++;
dev->commands_issued++;
LIST_INSERT_HEAD(&ahd->pending_scbs, scb, pending_links);
ahd_queue_scb(ahd, scb);
scb->platform_data->flags |= AHD_SCB_UP_EH_SEM;
spin_unlock_irq(&ahd->platform_data->spin_lock);
init_timer(&timer);
timer.data = (u_long)scb;
timer.expires = jiffies + (5 * HZ);
timer.function = ahd_linux_sem_timeout;
add_timer(&timer);
printf("Recovery code sleeping\n");
down(&ahd->platform_data->eh_sem);
printf("Recovery code awake\n");
retval = SUCCESS;
if (del_timer_sync(&timer) == 0) {
printf("Timer Expired\n");
retval = FAILED;
}
spin_lock_irq(&ahd->platform_data->spin_lock);
ahd_schedule_runq(ahd);
ahd_linux_run_complete_queue(ahd);
ahd_midlayer_entrypoint_unlock(ahd, &s);
printf("%s: Device reset returning 0x%x\n", ahd_name(ahd), retval);
return (retval);
}
/*
* Reset the SCSI bus.
*/
static int
ahd_linux_bus_reset(Scsi_Cmnd *cmd)
{
struct ahd_softc *ahd;
u_long s;
int found;
ahd = *(struct ahd_softc **)cmd->device->host->hostdata;
#ifdef AHD_DEBUG
if ((ahd_debug & AHD_SHOW_RECOVERY) != 0)
printf("%s: Bus reset called for cmd %p\n",
ahd_name(ahd), cmd);
#endif
ahd_midlayer_entrypoint_lock(ahd, &s);
found = ahd_reset_channel(ahd, cmd->device->channel + 'A',
/*initiate reset*/TRUE);
ahd_linux_run_complete_queue(ahd);
ahd_midlayer_entrypoint_unlock(ahd, &s);
if (bootverbose)
printf("%s: SCSI bus reset delivered. "
"%d SCBs aborted.\n", ahd_name(ahd), found);
return (SUCCESS);
}
Scsi_Host_Template aic79xx_driver_template = {
.module = THIS_MODULE,
.name = "aic79xx",
.proc_info = ahd_linux_proc_info,
.info = ahd_linux_info,
.queuecommand = ahd_linux_queue,
.eh_abort_handler = ahd_linux_abort,
.eh_device_reset_handler = ahd_linux_dev_reset,
.eh_bus_reset_handler = ahd_linux_bus_reset,
#if defined(__i386__)
.bios_param = ahd_linux_biosparam,
#endif
.can_queue = AHD_MAX_QUEUE,
.this_id = -1,
.cmd_per_lun = 2,
.use_clustering = ENABLE_CLUSTERING,
.slave_alloc = ahd_linux_slave_alloc,
.slave_configure = ahd_linux_slave_configure,
.slave_destroy = ahd_linux_slave_destroy,
};
/**************************** Tasklet Handler *********************************/
/*
* In 2.4.X and above, this routine is called from a tasklet,
* so we must re-acquire our lock prior to executing this code.
* In all prior kernels, ahd_schedule_runq() calls this routine
* directly and ahd_schedule_runq() is called with our lock held.
*/
static void
ahd_runq_tasklet(unsigned long data)
{
struct ahd_softc* ahd;
struct ahd_linux_device *dev;
u_long flags;
ahd = (struct ahd_softc *)data;
ahd_lock(ahd, &flags);
while ((dev = ahd_linux_next_device_to_run(ahd)) != NULL) {
TAILQ_REMOVE(&ahd->platform_data->device_runq, dev, links);
dev->flags &= ~AHD_DEV_ON_RUN_LIST;
ahd_linux_check_device_queue(ahd, dev);
/* Yeild to our interrupt handler */
ahd_unlock(ahd, &flags);
ahd_lock(ahd, &flags);
}
ahd_unlock(ahd, &flags);
}
/******************************** Bus DMA *************************************/
int
ahd_dma_tag_create(struct ahd_softc *ahd, bus_dma_tag_t parent,
bus_size_t alignment, bus_size_t boundary,
dma_addr_t lowaddr, dma_addr_t highaddr,
bus_dma_filter_t *filter, void *filterarg,
bus_size_t maxsize, int nsegments,
bus_size_t maxsegsz, int flags, bus_dma_tag_t *ret_tag)
{
bus_dma_tag_t dmat;
dmat = malloc(sizeof(*dmat), M_DEVBUF, M_NOWAIT);
if (dmat == NULL)
return (ENOMEM);
/*
* Linux is very simplistic about DMA memory. For now don't
* maintain all specification information. Once Linux supplies
* better facilities for doing these operations, or the
* needs of this particular driver change, we might need to do
* more here.
*/
dmat->alignment = alignment;
dmat->boundary = boundary;
dmat->maxsize = maxsize;
*ret_tag = dmat;
return (0);
}
void
ahd_dma_tag_destroy(struct ahd_softc *ahd, bus_dma_tag_t dmat)
{
free(dmat, M_DEVBUF);
}
int
ahd_dmamem_alloc(struct ahd_softc *ahd, bus_dma_tag_t dmat, void** vaddr,
int flags, bus_dmamap_t *mapp)
{
bus_dmamap_t map;
map = malloc(sizeof(*map), M_DEVBUF, M_NOWAIT);
if (map == NULL)
return (ENOMEM);
/*
* Although we can dma data above 4GB, our
* "consistent" memory is below 4GB for
* space efficiency reasons (only need a 4byte
* address). For this reason, we have to reset
* our dma mask when doing allocations.
*/
if (ahd->dev_softc != NULL)
if (pci_set_dma_mask(ahd->dev_softc, 0xFFFFFFFF)) {
printk(KERN_WARNING "aic79xx: No suitable DMA available.\n");
kfree(map);
return (ENODEV);
}
*vaddr = pci_alloc_consistent(ahd->dev_softc,
dmat->maxsize, &map->bus_addr);
if (ahd->dev_softc != NULL)
if (pci_set_dma_mask(ahd->dev_softc,
ahd->platform_data->hw_dma_mask)) {
printk(KERN_WARNING "aic79xx: No suitable DMA available.\n");
kfree(map);
return (ENODEV);
}
if (*vaddr == NULL)
return (ENOMEM);
*mapp = map;
return(0);
}
void
ahd_dmamem_free(struct ahd_softc *ahd, bus_dma_tag_t dmat,
void* vaddr, bus_dmamap_t map)
{
pci_free_consistent(ahd->dev_softc, dmat->maxsize,
vaddr, map->bus_addr);
}
int
ahd_dmamap_load(struct ahd_softc *ahd, bus_dma_tag_t dmat, bus_dmamap_t map,
void *buf, bus_size_t buflen, bus_dmamap_callback_t *cb,
void *cb_arg, int flags)
{
/*
* Assume for now that this will only be used during
* initialization and not for per-transaction buffer mapping.
*/
bus_dma_segment_t stack_sg;
stack_sg.ds_addr = map->bus_addr;
stack_sg.ds_len = dmat->maxsize;
cb(cb_arg, &stack_sg, /*nseg*/1, /*error*/0);
return (0);
}
void
ahd_dmamap_destroy(struct ahd_softc *ahd, bus_dma_tag_t dmat, bus_dmamap_t map)
{
/*
* The map may is NULL in our < 2.3.X implementation.
*/
if (map != NULL)
free(map, M_DEVBUF);
}
int
ahd_dmamap_unload(struct ahd_softc *ahd, bus_dma_tag_t dmat, bus_dmamap_t map)
{
/* Nothing to do */
return (0);
}
/********************* Platform Dependent Functions ***************************/
/*
* Compare "left hand" softc with "right hand" softc, returning:
* < 0 - lahd has a lower priority than rahd
* 0 - Softcs are equal
* > 0 - lahd has a higher priority than rahd
*/
int
ahd_softc_comp(struct ahd_softc *lahd, struct ahd_softc *rahd)
{
int value;
/*
* Under Linux, cards are ordered as follows:
* 1) PCI devices that are marked as the boot controller.
* 2) PCI devices with BIOS enabled sorted by bus/slot/func.
* 3) All remaining PCI devices sorted by bus/slot/func.
*/
#if 0
value = (lahd->flags & AHD_BOOT_CHANNEL)
- (rahd->flags & AHD_BOOT_CHANNEL);
if (value != 0)
/* Controllers set for boot have a *higher* priority */
return (value);
#endif
value = (lahd->flags & AHD_BIOS_ENABLED)
- (rahd->flags & AHD_BIOS_ENABLED);
if (value != 0)
/* Controllers with BIOS enabled have a *higher* priority */
return (value);
/* Still equal. Sort by bus/slot/func. */
if (aic79xx_reverse_scan != 0)
value = ahd_get_pci_bus(lahd->dev_softc)
- ahd_get_pci_bus(rahd->dev_softc);
else
value = ahd_get_pci_bus(rahd->dev_softc)
- ahd_get_pci_bus(lahd->dev_softc);
if (value != 0)
return (value);
if (aic79xx_reverse_scan != 0)
value = ahd_get_pci_slot(lahd->dev_softc)
- ahd_get_pci_slot(rahd->dev_softc);
else
value = ahd_get_pci_slot(rahd->dev_softc)
- ahd_get_pci_slot(lahd->dev_softc);
if (value != 0)
return (value);
value = rahd->channel - lahd->channel;
return (value);
}
static void
ahd_linux_setup_tag_info(u_long arg, int instance, int targ, int32_t value)
{
if ((instance >= 0) && (targ >= 0)
&& (instance < NUM_ELEMENTS(aic79xx_tag_info))
&& (targ < AHD_NUM_TARGETS)) {
aic79xx_tag_info[instance].tag_commands[targ] = value & 0x1FF;
if (bootverbose)
printf("tag_info[%d:%d] = %d\n", instance, targ, value);
}
}
static void
ahd_linux_setup_rd_strm_info(u_long arg, int instance, int targ, int32_t value)
{
if ((instance >= 0)
&& (instance < NUM_ELEMENTS(aic79xx_rd_strm_info))) {
aic79xx_rd_strm_info[instance] = value & 0xFFFF;
if (bootverbose)
printf("rd_strm[%d] = 0x%x\n", instance, value);
}
}
static void
ahd_linux_setup_dv(u_long arg, int instance, int targ, int32_t value)
{
if ((instance >= 0)
&& (instance < NUM_ELEMENTS(aic79xx_dv_settings))) {
aic79xx_dv_settings[instance] = value;
if (bootverbose)
printf("dv[%d] = %d\n", instance, value);
}
}
static void
ahd_linux_setup_iocell_info(u_long index, int instance, int targ, int32_t value)
{
if ((instance >= 0)
&& (instance < NUM_ELEMENTS(aic79xx_iocell_info))) {
uint8_t *iocell_info;
iocell_info = (uint8_t*)&aic79xx_iocell_info[instance];
iocell_info[index] = value & 0xFFFF;
if (bootverbose)
printf("iocell[%d:%ld] = %d\n", instance, index, value);
}
}
static void
ahd_linux_setup_tag_info_global(char *p)
{
int tags, i, j;
tags = simple_strtoul(p + 1, NULL, 0) & 0xff;
printf("Setting Global Tags= %d\n", tags);
for (i = 0; i < NUM_ELEMENTS(aic79xx_tag_info); i++) {
for (j = 0; j < AHD_NUM_TARGETS; j++) {
aic79xx_tag_info[i].tag_commands[j] = tags;
}
}
}
/*
* Handle Linux boot parameters. This routine allows for assigning a value
* to a parameter with a ':' between the parameter and the value.
* ie. aic79xx=stpwlev:1,extended
*/
static int
aic79xx_setup(char *s)
{
int i, n;
char *p;
char *end;
static struct {
const char *name;
uint32_t *flag;
} options[] = {
{ "extended", &aic79xx_extended },
{ "no_reset", &aic79xx_no_reset },
{ "verbose", &aic79xx_verbose },
{ "allow_memio", &aic79xx_allow_memio},
#ifdef AHD_DEBUG
{ "debug", &ahd_debug },
#endif
{ "reverse_scan", &aic79xx_reverse_scan },
{ "periodic_otag", &aic79xx_periodic_otag },
{ "pci_parity", &aic79xx_pci_parity },
{ "seltime", &aic79xx_seltime },
{ "tag_info", NULL },
{ "global_tag_depth", NULL},
{ "rd_strm", NULL },
{ "dv", NULL },
{ "slewrate", NULL },
{ "precomp", NULL },
{ "amplitude", NULL },
};
end = strchr(s, '\0');
/*
* XXX ia64 gcc isn't smart enough to know that NUM_ELEMENTS
* will never be 0 in this case.
*/
n = 0;
while ((p = strsep(&s, ",.")) != NULL) {
if (*p == '\0')
continue;
for (i = 0; i < NUM_ELEMENTS(options); i++) {
n = strlen(options[i].name);
if (strncmp(options[i].name, p, n) == 0)
break;
}
if (i == NUM_ELEMENTS(options))
continue;
if (strncmp(p, "global_tag_depth", n) == 0) {
ahd_linux_setup_tag_info_global(p + n);
} else if (strncmp(p, "tag_info", n) == 0) {
s = aic_parse_brace_option("tag_info", p + n, end,
2, ahd_linux_setup_tag_info, 0);
} else if (strncmp(p, "rd_strm", n) == 0) {
s = aic_parse_brace_option("rd_strm", p + n, end,
1, ahd_linux_setup_rd_strm_info, 0);
} else if (strncmp(p, "dv", n) == 0) {
s = aic_parse_brace_option("dv", p + n, end, 1,
ahd_linux_setup_dv, 0);
} else if (strncmp(p, "slewrate", n) == 0) {
s = aic_parse_brace_option("slewrate",
p + n, end, 1, ahd_linux_setup_iocell_info,
AIC79XX_SLEWRATE_INDEX);
} else if (strncmp(p, "precomp", n) == 0) {
s = aic_parse_brace_option("precomp",
p + n, end, 1, ahd_linux_setup_iocell_info,
AIC79XX_PRECOMP_INDEX);
} else if (strncmp(p, "amplitude", n) == 0) {
s = aic_parse_brace_option("amplitude",
p + n, end, 1, ahd_linux_setup_iocell_info,
AIC79XX_AMPLITUDE_INDEX);
} else if (p[n] == ':') {
*(options[i].flag) = simple_strtoul(p + n + 1, NULL, 0);
} else if (!strncmp(p, "verbose", n)) {
*(options[i].flag) = 1;
} else {
*(options[i].flag) ^= 0xFFFFFFFF;
}
}
return 1;
}
__setup("aic79xx=", aic79xx_setup);
uint32_t aic79xx_verbose;
int
ahd_linux_register_host(struct ahd_softc *ahd, Scsi_Host_Template *template)
{
char buf[80];
struct Scsi_Host *host;
char *new_name;
u_long s;
u_long target;
template->name = ahd->description;
host = scsi_host_alloc(template, sizeof(struct ahd_softc *));
if (host == NULL)
return (ENOMEM);
*((struct ahd_softc **)host->hostdata) = ahd;
ahd_lock(ahd, &s);
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
scsi_assign_lock(host, &ahd->platform_data->spin_lock);
#elif AHD_SCSI_HAS_HOST_LOCK != 0
host->lock = &ahd->platform_data->spin_lock;
#endif
ahd->platform_data->host = host;
host->can_queue = AHD_MAX_QUEUE;
host->cmd_per_lun = 2;
host->sg_tablesize = AHD_NSEG;
host->this_id = ahd->our_id;
host->irq = ahd->platform_data->irq;
host->max_id = (ahd->features & AHD_WIDE) ? 16 : 8;
host->max_lun = AHD_NUM_LUNS;
host->max_channel = 0;
host->sg_tablesize = AHD_NSEG;
ahd_set_unit(ahd, ahd_linux_next_unit());
sprintf(buf, "scsi%d", host->host_no);
new_name = malloc(strlen(buf) + 1, M_DEVBUF, M_NOWAIT);
if (new_name != NULL) {
strcpy(new_name, buf);
ahd_set_name(ahd, new_name);
}
host->unique_id = ahd->unit;
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
scsi_set_pci_device(host, ahd->dev_softc);
#endif
ahd_linux_setup_user_rd_strm_settings(ahd);
ahd_linux_initialize_scsi_bus(ahd);
ahd_unlock(ahd, &s);
ahd->platform_data->dv_pid = kernel_thread(ahd_linux_dv_thread, ahd, 0);
ahd_lock(ahd, &s);
if (ahd->platform_data->dv_pid < 0) {
printf("%s: Failed to create DV thread, error= %d\n",
ahd_name(ahd), ahd->platform_data->dv_pid);
return (-ahd->platform_data->dv_pid);
}
/*
* Initially allocate *all* of our linux target objects
* so that the DV thread will scan them all in parallel
* just after driver initialization. Any device that
* does not exist will have its target object destroyed
* by the selection timeout handler. In the case of a
* device that appears after the initial DV scan, async
* negotiation will occur for the first command, and DV
* will comence should that first command be successful.
*/
for (target = 0; target < host->max_id; target++) {
/*
* Skip our own ID. Some Compaq/HP storage devices
* have enclosure management devices that respond to
* single bit selection (i.e. selecting ourselves).
* It is expected that either an external application
* or a modified kernel will be used to probe this
* ID if it is appropriate. To accommodate these
* installations, ahc_linux_alloc_target() will allocate
* for our ID if asked to do so.
*/
if (target == ahd->our_id)
continue;
ahd_linux_alloc_target(ahd, 0, target);
}
ahd_intr_enable(ahd, TRUE);
ahd_linux_start_dv(ahd);
ahd_unlock(ahd, &s);
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
scsi_add_host(host, &ahd->dev_softc->dev); /* XXX handle failure */
scsi_scan_host(host);
#endif
return (0);
}
uint64_t
ahd_linux_get_memsize(void)
{
struct sysinfo si;
si_meminfo(&si);
return ((uint64_t)si.totalram << PAGE_SHIFT);
}
/*
* Find the smallest available unit number to use
* for a new device. We don't just use a static
* count to handle the "repeated hot-(un)plug"
* scenario.
*/
static int
ahd_linux_next_unit(void)
{
struct ahd_softc *ahd;
int unit;
unit = 0;
retry:
TAILQ_FOREACH(ahd, &ahd_tailq, links) {
if (ahd->unit == unit) {
unit++;
goto retry;
}
}
return (unit);
}
/*
* Place the SCSI bus into a known state by either resetting it,
* or forcing transfer negotiations on the next command to any
* target.
*/
static void
ahd_linux_initialize_scsi_bus(struct ahd_softc *ahd)
{
u_int target_id;
u_int numtarg;
target_id = 0;
numtarg = 0;
if (aic79xx_no_reset != 0)
ahd->flags &= ~AHD_RESET_BUS_A;
if ((ahd->flags & AHD_RESET_BUS_A) != 0)
ahd_reset_channel(ahd, 'A', /*initiate_reset*/TRUE);
else
numtarg = (ahd->features & AHD_WIDE) ? 16 : 8;
/*
* Force negotiation to async for all targets that
* will not see an initial bus reset.
*/
for (; target_id < numtarg; target_id++) {
struct ahd_devinfo devinfo;
struct ahd_initiator_tinfo *tinfo;
struct ahd_tmode_tstate *tstate;
tinfo = ahd_fetch_transinfo(ahd, 'A', ahd->our_id,
target_id, &tstate);
ahd_compile_devinfo(&devinfo, ahd->our_id, target_id,
CAM_LUN_WILDCARD, 'A', ROLE_INITIATOR);
ahd_update_neg_request(ahd, &devinfo, tstate,
tinfo, AHD_NEG_ALWAYS);
}
/* Give the bus some time to recover */
if ((ahd->flags & AHD_RESET_BUS_A) != 0) {
ahd_freeze_simq(ahd);
init_timer(&ahd->platform_data->reset_timer);
ahd->platform_data->reset_timer.data = (u_long)ahd;
ahd->platform_data->reset_timer.expires =
jiffies + (AIC79XX_RESET_DELAY * HZ)/1000;
ahd->platform_data->reset_timer.function =
(ahd_linux_callback_t *)ahd_release_simq;
add_timer(&ahd->platform_data->reset_timer);
}
}
int
ahd_platform_alloc(struct ahd_softc *ahd, void *platform_arg)
{
ahd->platform_data =
malloc(sizeof(struct ahd_platform_data), M_DEVBUF, M_NOWAIT);
if (ahd->platform_data == NULL)
return (ENOMEM);
memset(ahd->platform_data, 0, sizeof(struct ahd_platform_data));
TAILQ_INIT(&ahd->platform_data->completeq);
TAILQ_INIT(&ahd->platform_data->device_runq);
ahd->platform_data->irq = AHD_LINUX_NOIRQ;
ahd->platform_data->hw_dma_mask = 0xFFFFFFFF;
ahd_lockinit(ahd);
ahd_done_lockinit(ahd);
init_timer(&ahd->platform_data->completeq_timer);
ahd->platform_data->completeq_timer.data = (u_long)ahd;
ahd->platform_data->completeq_timer.function =
(ahd_linux_callback_t *)ahd_linux_thread_run_complete_queue;
init_MUTEX_LOCKED(&ahd->platform_data->eh_sem);
init_MUTEX_LOCKED(&ahd->platform_data->dv_sem);
init_MUTEX_LOCKED(&ahd->platform_data->dv_cmd_sem);
ahd_setup_runq_tasklet(ahd);
ahd->seltime = (aic79xx_seltime & 0x3) << 4;
return (0);
}
void
ahd_platform_free(struct ahd_softc *ahd)
{
struct ahd_linux_target *targ;
struct ahd_linux_device *dev;
int i, j;
if (ahd->platform_data != NULL) {
del_timer_sync(&ahd->platform_data->completeq_timer);
ahd_linux_kill_dv_thread(ahd);
ahd_teardown_runq_tasklet(ahd);
if (ahd->platform_data->host != NULL) {
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
scsi_remove_host(ahd->platform_data->host);
#endif
scsi_host_put(ahd->platform_data->host);
}
/* destroy all of the device and target objects */
for (i = 0; i < AHD_NUM_TARGETS; i++) {
targ = ahd->platform_data->targets[i];
if (targ != NULL) {
/* Keep target around through the loop. */
targ->refcount++;
for (j = 0; j < AHD_NUM_LUNS; j++) {
if (targ->devices[j] == NULL)
continue;
dev = targ->devices[j];
ahd_linux_free_device(ahd, dev);
}
/*
* Forcibly free the target now that
* all devices are gone.
*/
ahd_linux_free_target(ahd, targ);
}
}
if (ahd->platform_data->irq != AHD_LINUX_NOIRQ)
free_irq(ahd->platform_data->irq, ahd);
if (ahd->tags[0] == BUS_SPACE_PIO
&& ahd->bshs[0].ioport != 0)
release_region(ahd->bshs[0].ioport, 256);
if (ahd->tags[1] == BUS_SPACE_PIO
&& ahd->bshs[1].ioport != 0)
release_region(ahd->bshs[1].ioport, 256);
if (ahd->tags[0] == BUS_SPACE_MEMIO
&& ahd->bshs[0].maddr != NULL) {
iounmap(ahd->bshs[0].maddr);
release_mem_region(ahd->platform_data->mem_busaddr,
0x1000);
}
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
/*
* In 2.4 we detach from the scsi midlayer before the PCI
* layer invokes our remove callback. No per-instance
* detach is provided, so we must reach inside the PCI
* subsystem's internals and detach our driver manually.
*/
if (ahd->dev_softc != NULL)
ahd->dev_softc->driver = NULL;
#endif
free(ahd->platform_data, M_DEVBUF);
}
}
void
ahd_platform_init(struct ahd_softc *ahd)
{
/*
* Lookup and commit any modified IO Cell options.
*/
if (ahd->unit < NUM_ELEMENTS(aic79xx_iocell_info)) {
struct ahd_linux_iocell_opts *iocell_opts;
iocell_opts = &aic79xx_iocell_info[ahd->unit];
if (iocell_opts->precomp != AIC79XX_DEFAULT_PRECOMP)
AHD_SET_PRECOMP(ahd, iocell_opts->precomp);
if (iocell_opts->slewrate != AIC79XX_DEFAULT_SLEWRATE)
AHD_SET_SLEWRATE(ahd, iocell_opts->slewrate);
if (iocell_opts->amplitude != AIC79XX_DEFAULT_AMPLITUDE)
AHD_SET_AMPLITUDE(ahd, iocell_opts->amplitude);
}
}
void
ahd_platform_freeze_devq(struct ahd_softc *ahd, struct scb *scb)
{
ahd_platform_abort_scbs(ahd, SCB_GET_TARGET(ahd, scb),
SCB_GET_CHANNEL(ahd, scb),
SCB_GET_LUN(scb), SCB_LIST_NULL,
ROLE_UNKNOWN, CAM_REQUEUE_REQ);
}
void
ahd_platform_set_tags(struct ahd_softc *ahd, struct ahd_devinfo *devinfo,
ahd_queue_alg alg)
{
struct ahd_linux_device *dev;
int was_queuing;
int now_queuing;
dev = ahd_linux_get_device(ahd, devinfo->channel - 'A',
devinfo->target,
devinfo->lun, /*alloc*/FALSE);
if (dev == NULL)
return;
was_queuing = dev->flags & (AHD_DEV_Q_BASIC|AHD_DEV_Q_TAGGED);
switch (alg) {
default:
case AHD_QUEUE_NONE:
now_queuing = 0;
break;
case AHD_QUEUE_BASIC:
now_queuing = AHD_DEV_Q_BASIC;
break;
case AHD_QUEUE_TAGGED:
now_queuing = AHD_DEV_Q_TAGGED;
break;
}
if ((dev->flags & AHD_DEV_FREEZE_TIL_EMPTY) == 0
&& (was_queuing != now_queuing)
&& (dev->active != 0)) {
dev->flags |= AHD_DEV_FREEZE_TIL_EMPTY;
dev->qfrozen++;
}
dev->flags &= ~(AHD_DEV_Q_BASIC|AHD_DEV_Q_TAGGED|AHD_DEV_PERIODIC_OTAG);
if (now_queuing) {
u_int usertags;
usertags = ahd_linux_user_tagdepth(ahd, devinfo);
if (!was_queuing) {
/*
* Start out agressively and allow our
* dynamic queue depth algorithm to take
* care of the rest.
*/
dev->maxtags = usertags;
dev->openings = dev->maxtags - dev->active;
}
if (dev->maxtags == 0) {
/*
* Queueing is disabled by the user.
*/
dev->openings = 1;
} else if (alg == AHD_QUEUE_TAGGED) {
dev->flags |= AHD_DEV_Q_TAGGED;
if (aic79xx_periodic_otag != 0)
dev->flags |= AHD_DEV_PERIODIC_OTAG;
} else
dev->flags |= AHD_DEV_Q_BASIC;
} else {
/* We can only have one opening. */
dev->maxtags = 0;
dev->openings = 1 - dev->active;
}
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
if (dev->scsi_device != NULL) {
switch ((dev->flags & (AHD_DEV_Q_BASIC|AHD_DEV_Q_TAGGED))) {
case AHD_DEV_Q_BASIC:
scsi_adjust_queue_depth(dev->scsi_device,
MSG_SIMPLE_TASK,
dev->openings + dev->active);
break;
case AHD_DEV_Q_TAGGED:
scsi_adjust_queue_depth(dev->scsi_device,
MSG_ORDERED_TASK,
dev->openings + dev->active);
break;
default:
/*
* We allow the OS to queue 2 untagged transactions to
* us at any time even though we can only execute them
* serially on the controller/device. This should
* remove some latency.
*/
scsi_adjust_queue_depth(dev->scsi_device,
/*NON-TAGGED*/0,
/*queue depth*/2);
break;
}
}
#endif
}
int
ahd_platform_abort_scbs(struct ahd_softc *ahd, int target, char channel,
int lun, u_int tag, role_t role, uint32_t status)
{
int targ;
int maxtarg;
int maxlun;
int clun;
int count;
if (tag != SCB_LIST_NULL)
return (0);
targ = 0;
if (target != CAM_TARGET_WILDCARD) {
targ = target;
maxtarg = targ + 1;
} else {
maxtarg = (ahd->features & AHD_WIDE) ? 16 : 8;
}
clun = 0;
if (lun != CAM_LUN_WILDCARD) {
clun = lun;
maxlun = clun + 1;
} else {
maxlun = AHD_NUM_LUNS;
}
count = 0;
for (; targ < maxtarg; targ++) {
for (; clun < maxlun; clun++) {
struct ahd_linux_device *dev;
struct ahd_busyq *busyq;
struct ahd_cmd *acmd;
dev = ahd_linux_get_device(ahd, /*chan*/0, targ,
clun, /*alloc*/FALSE);
if (dev == NULL)
continue;
busyq = &dev->busyq;
while ((acmd = TAILQ_FIRST(busyq)) != NULL) {
Scsi_Cmnd *cmd;
cmd = &acmd_scsi_cmd(acmd);
TAILQ_REMOVE(busyq, acmd,
acmd_links.tqe);
count++;
cmd->result = status << 16;
ahd_linux_queue_cmd_complete(ahd, cmd);
}
}
}
return (count);
}
static void
ahd_linux_thread_run_complete_queue(struct ahd_softc *ahd)
{
u_long flags;
ahd_lock(ahd, &flags);
del_timer(&ahd->platform_data->completeq_timer);
ahd->platform_data->flags &= ~AHD_RUN_CMPLT_Q_TIMER;
ahd_linux_run_complete_queue(ahd);
ahd_unlock(ahd, &flags);
}
static void
ahd_linux_start_dv(struct ahd_softc *ahd)
{
/*
* Freeze the simq and signal ahd_linux_queue to not let any
* more commands through
*/
if ((ahd->platform_data->flags & AHD_DV_ACTIVE) == 0) {
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV)
printf("%s: Starting DV\n", ahd_name(ahd));
#endif
ahd->platform_data->flags |= AHD_DV_ACTIVE;
ahd_freeze_simq(ahd);
/* Wake up the DV kthread */
up(&ahd->platform_data->dv_sem);
}
}
static int
ahd_linux_dv_thread(void *data)
{
struct ahd_softc *ahd;
int target;
u_long s;
ahd = (struct ahd_softc *)data;
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV)
printf("In DV Thread\n");
#endif
/*
* Complete thread creation.
*/
lock_kernel();
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,60)
/*
* Don't care about any signals.
*/
siginitsetinv(&current->blocked, 0);
daemonize();
sprintf(current->comm, "ahd_dv_%d", ahd->unit);
#else
daemonize("ahd_dv_%d", ahd->unit);
current->flags |= PF_FREEZE;
#endif
unlock_kernel();
while (1) {
/*
* Use down_interruptible() rather than down() to
* avoid inclusion in the load average.
*/
down_interruptible(&ahd->platform_data->dv_sem);
/* Check to see if we've been signaled to exit */
ahd_lock(ahd, &s);
if ((ahd->platform_data->flags & AHD_DV_SHUTDOWN) != 0) {
ahd_unlock(ahd, &s);
break;
}
ahd_unlock(ahd, &s);
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV)
printf("%s: Beginning Domain Validation\n",
ahd_name(ahd));
#endif
/*
* Wait for any pending commands to drain before proceeding.
*/
ahd_lock(ahd, &s);
while (LIST_FIRST(&ahd->pending_scbs) != NULL) {
ahd->platform_data->flags |= AHD_DV_WAIT_SIMQ_EMPTY;
ahd_unlock(ahd, &s);
down_interruptible(&ahd->platform_data->dv_sem);
ahd_lock(ahd, &s);
}
/*
* Wait for the SIMQ to be released so that DV is the
* only reason the queue is frozen.
*/
while (AHD_DV_SIMQ_FROZEN(ahd) == 0) {
ahd->platform_data->flags |= AHD_DV_WAIT_SIMQ_RELEASE;
ahd_unlock(ahd, &s);
down_interruptible(&ahd->platform_data->dv_sem);
ahd_lock(ahd, &s);
}
ahd_unlock(ahd, &s);
for (target = 0; target < AHD_NUM_TARGETS; target++)
ahd_linux_dv_target(ahd, target);
ahd_lock(ahd, &s);
ahd->platform_data->flags &= ~AHD_DV_ACTIVE;
ahd_unlock(ahd, &s);
/*
* Release the SIMQ so that normal commands are
* allowed to continue on the bus.
*/
ahd_release_simq(ahd);
}
up(&ahd->platform_data->eh_sem);
return (0);
}
static void
ahd_linux_kill_dv_thread(struct ahd_softc *ahd)
{
u_long s;
ahd_lock(ahd, &s);
if (ahd->platform_data->dv_pid != 0) {
ahd->platform_data->flags |= AHD_DV_SHUTDOWN;
ahd_unlock(ahd, &s);
up(&ahd->platform_data->dv_sem);
/*
* Use the eh_sem as an indicator that the
* dv thread is exiting. Note that the dv
* thread must still return after performing
* the up on our semaphore before it has
* completely exited this module. Unfortunately,
* there seems to be no easy way to wait for the
* exit of a thread for which you are not the
* parent (dv threads are parented by init).
* Cross your fingers...
*/
down(&ahd->platform_data->eh_sem);
/*
* Mark the dv thread as already dead. This
* avoids attempting to kill it a second time.
* This is necessary because we must kill the
* DV thread before calling ahd_free() in the
* module shutdown case to avoid bogus locking
* in the SCSI mid-layer, but we ahd_free() is
* called without killing the DV thread in the
* instance detach case, so ahd_platform_free()
* calls us again to verify that the DV thread
* is dead.
*/
ahd->platform_data->dv_pid = 0;
} else {
ahd_unlock(ahd, &s);
}
}
#define AHD_LINUX_DV_INQ_SHORT_LEN 36
#define AHD_LINUX_DV_INQ_LEN 256
#define AHD_LINUX_DV_TIMEOUT (HZ / 4)
#define AHD_SET_DV_STATE(ahd, targ, newstate) \
ahd_set_dv_state(ahd, targ, newstate, __LINE__)
static __inline void
ahd_set_dv_state(struct ahd_softc *ahd, struct ahd_linux_target *targ,
ahd_dv_state newstate, u_int line)
{
ahd_dv_state oldstate;
oldstate = targ->dv_state;
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV)
printf("%s:%d: Going from state %d to state %d\n",
ahd_name(ahd), line, oldstate, newstate);
#endif
if (oldstate == newstate)
targ->dv_state_retry++;
else
targ->dv_state_retry = 0;
targ->dv_state = newstate;
}
static void
ahd_linux_dv_target(struct ahd_softc *ahd, u_int target_offset)
{
struct ahd_devinfo devinfo;
struct ahd_linux_target *targ;
struct scsi_cmnd *cmd;
struct scsi_device *scsi_dev;
struct scsi_sense_data *sense;
uint8_t *buffer;
u_long s;
u_int timeout;
int echo_size;
sense = NULL;
buffer = NULL;
echo_size = 0;
ahd_lock(ahd, &s);
targ = ahd->platform_data->targets[target_offset];
if (targ == NULL || (targ->flags & AHD_DV_REQUIRED) == 0) {
ahd_unlock(ahd, &s);
return;
}
ahd_compile_devinfo(&devinfo, ahd->our_id, targ->target, /*lun*/0,
targ->channel + 'A', ROLE_INITIATOR);
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, &devinfo);
printf("Performing DV\n");
}
#endif
ahd_unlock(ahd, &s);
cmd = malloc(sizeof(struct scsi_cmnd), M_DEVBUF, M_WAITOK);
scsi_dev = malloc(sizeof(struct scsi_device), M_DEVBUF, M_WAITOK);
scsi_dev->host = ahd->platform_data->host;
scsi_dev->id = devinfo.target;
scsi_dev->lun = devinfo.lun;
scsi_dev->channel = devinfo.channel - 'A';
ahd->platform_data->dv_scsi_dev = scsi_dev;
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_INQ_SHORT_ASYNC);
while (targ->dv_state != AHD_DV_STATE_EXIT) {
timeout = AHD_LINUX_DV_TIMEOUT;
switch (targ->dv_state) {
case AHD_DV_STATE_INQ_SHORT_ASYNC:
case AHD_DV_STATE_INQ_ASYNC:
case AHD_DV_STATE_INQ_ASYNC_VERIFY:
/*
* Set things to async narrow to reduce the
* chance that the INQ will fail.
*/
ahd_lock(ahd, &s);
ahd_set_syncrate(ahd, &devinfo, 0, 0, 0,
AHD_TRANS_GOAL, /*paused*/FALSE);
ahd_set_width(ahd, &devinfo, MSG_EXT_WDTR_BUS_8_BIT,
AHD_TRANS_GOAL, /*paused*/FALSE);
ahd_unlock(ahd, &s);
timeout = 10 * HZ;
targ->flags &= ~AHD_INQ_VALID;
/* FALLTHROUGH */
case AHD_DV_STATE_INQ_VERIFY:
{
u_int inq_len;
if (targ->dv_state == AHD_DV_STATE_INQ_SHORT_ASYNC)
inq_len = AHD_LINUX_DV_INQ_SHORT_LEN;
else
inq_len = targ->inq_data->additional_length + 5;
ahd_linux_dv_inq(ahd, cmd, &devinfo, targ, inq_len);
break;
}
case AHD_DV_STATE_TUR:
case AHD_DV_STATE_BUSY:
timeout = 5 * HZ;
ahd_linux_dv_tur(ahd, cmd, &devinfo);
break;
case AHD_DV_STATE_REBD:
ahd_linux_dv_rebd(ahd, cmd, &devinfo, targ);
break;
case AHD_DV_STATE_WEB:
ahd_linux_dv_web(ahd, cmd, &devinfo, targ);
break;
case AHD_DV_STATE_REB:
ahd_linux_dv_reb(ahd, cmd, &devinfo, targ);
break;
case AHD_DV_STATE_SU:
ahd_linux_dv_su(ahd, cmd, &devinfo, targ);
timeout = 50 * HZ;
break;
default:
ahd_print_devinfo(ahd, &devinfo);
printf("Unknown DV state %d\n", targ->dv_state);
goto out;
}
/* Queue the command and wait for it to complete */
/* Abuse eh_timeout in the scsi_cmnd struct for our purposes */
init_timer(&cmd->eh_timeout);
#ifdef AHD_DEBUG
if ((ahd_debug & AHD_SHOW_MESSAGES) != 0)
/*
* All of the printfs during negotiation
* really slow down the negotiation.
* Add a bit of time just to be safe.
*/
timeout += HZ;
#endif
scsi_add_timer(cmd, timeout, ahd_linux_dv_timeout);
/*
* In 2.5.X, it is assumed that all calls from the
* "midlayer" (which we are emulating) will have the
* ahd host lock held. For other kernels, the
* io_request_lock must be held.
*/
#if AHD_SCSI_HAS_HOST_LOCK != 0
ahd_lock(ahd, &s);
#else
spin_lock_irqsave(&io_request_lock, s);
#endif
ahd_linux_queue(cmd, ahd_linux_dv_complete);
#if AHD_SCSI_HAS_HOST_LOCK != 0
ahd_unlock(ahd, &s);
#else
spin_unlock_irqrestore(&io_request_lock, s);
#endif
down_interruptible(&ahd->platform_data->dv_cmd_sem);
/*
* Wait for the SIMQ to be released so that DV is the
* only reason the queue is frozen.
*/
ahd_lock(ahd, &s);
while (AHD_DV_SIMQ_FROZEN(ahd) == 0) {
ahd->platform_data->flags |= AHD_DV_WAIT_SIMQ_RELEASE;
ahd_unlock(ahd, &s);
down_interruptible(&ahd->platform_data->dv_sem);
ahd_lock(ahd, &s);
}
ahd_unlock(ahd, &s);
ahd_linux_dv_transition(ahd, cmd, &devinfo, targ);
}
out:
if ((targ->flags & AHD_INQ_VALID) != 0
&& ahd_linux_get_device(ahd, devinfo.channel - 'A',
devinfo.target, devinfo.lun,
/*alloc*/FALSE) == NULL) {
/*
* The DV state machine failed to configure this device.
* This is normal if DV is disabled. Since we have inquiry
* data, filter it and use the "optimistic" negotiation
* parameters found in the inquiry string.
*/
ahd_linux_filter_inquiry(ahd, &devinfo);
if ((targ->flags & (AHD_BASIC_DV|AHD_ENHANCED_DV)) != 0) {
ahd_print_devinfo(ahd, &devinfo);
printf("DV failed to configure device. "
"Please file a bug report against "
"this driver.\n");
}
}
if (cmd != NULL)
free(cmd, M_DEVBUF);
if (ahd->platform_data->dv_scsi_dev != NULL) {
free(ahd->platform_data->dv_scsi_dev, M_DEVBUF);
ahd->platform_data->dv_scsi_dev = NULL;
}
ahd_lock(ahd, &s);
if (targ->dv_buffer != NULL) {
free(targ->dv_buffer, M_DEVBUF);
targ->dv_buffer = NULL;
}
if (targ->dv_buffer1 != NULL) {
free(targ->dv_buffer1, M_DEVBUF);
targ->dv_buffer1 = NULL;
}
targ->flags &= ~AHD_DV_REQUIRED;
if (targ->refcount == 0)
ahd_linux_free_target(ahd, targ);
ahd_unlock(ahd, &s);
}
static __inline int
ahd_linux_dv_fallback(struct ahd_softc *ahd, struct ahd_devinfo *devinfo)
{
u_long s;
int retval;
ahd_lock(ahd, &s);
retval = ahd_linux_fallback(ahd, devinfo);
ahd_unlock(ahd, &s);
return (retval);
}
static void
ahd_linux_dv_transition(struct ahd_softc *ahd, struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo,
struct ahd_linux_target *targ)
{
u_int32_t status;
status = aic_error_action(cmd, targ->inq_data,
ahd_cmd_get_transaction_status(cmd),
ahd_cmd_get_scsi_status(cmd));
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Entering ahd_linux_dv_transition, state= %d, "
"status= 0x%x, cmd->result= 0x%x\n", targ->dv_state,
status, cmd->result);
}
#endif
switch (targ->dv_state) {
case AHD_DV_STATE_INQ_SHORT_ASYNC:
case AHD_DV_STATE_INQ_ASYNC:
switch (status & SS_MASK) {
case SS_NOP:
{
AHD_SET_DV_STATE(ahd, targ, targ->dv_state+1);
break;
}
case SS_INQ_REFRESH:
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
case SS_TUR:
case SS_RETRY:
AHD_SET_DV_STATE(ahd, targ, targ->dv_state);
if (ahd_cmd_get_transaction_status(cmd)
== CAM_REQUEUE_REQ)
targ->dv_state_retry--;
if ((status & SS_ERRMASK) == EBUSY)
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_BUSY);
if (targ->dv_state_retry < 10)
break;
/* FALLTHROUGH */
default:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Failed DV inquiry, skipping\n");
}
#endif
break;
}
break;
case AHD_DV_STATE_INQ_ASYNC_VERIFY:
switch (status & SS_MASK) {
case SS_NOP:
{
u_int xportflags;
u_int spi3data;
if (memcmp(targ->inq_data, targ->dv_buffer,
AHD_LINUX_DV_INQ_LEN) != 0) {
/*
* Inquiry data must have changed.
* Try from the top again.
*/
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
}
AHD_SET_DV_STATE(ahd, targ, targ->dv_state+1);
targ->flags |= AHD_INQ_VALID;
if (ahd_linux_user_dv_setting(ahd) == 0)
break;
xportflags = targ->inq_data->flags;
if ((xportflags & (SID_Sync|SID_WBus16)) == 0)
break;
spi3data = targ->inq_data->spi3data;
switch (spi3data & SID_SPI_CLOCK_DT_ST) {
default:
case SID_SPI_CLOCK_ST:
/* Assume only basic DV is supported. */
targ->flags |= AHD_BASIC_DV;
break;
case SID_SPI_CLOCK_DT:
case SID_SPI_CLOCK_DT_ST:
targ->flags |= AHD_ENHANCED_DV;
break;
}
break;
}
case SS_INQ_REFRESH:
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
case SS_TUR:
case SS_RETRY:
AHD_SET_DV_STATE(ahd, targ, targ->dv_state);
if (ahd_cmd_get_transaction_status(cmd)
== CAM_REQUEUE_REQ)
targ->dv_state_retry--;
if ((status & SS_ERRMASK) == EBUSY)
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_BUSY);
if (targ->dv_state_retry < 10)
break;
/* FALLTHROUGH */
default:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Failed DV inquiry, skipping\n");
}
#endif
break;
}
break;
case AHD_DV_STATE_INQ_VERIFY:
switch (status & SS_MASK) {
case SS_NOP:
{
if (memcmp(targ->inq_data, targ->dv_buffer,
AHD_LINUX_DV_INQ_LEN) == 0) {
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
int i;
ahd_print_devinfo(ahd, devinfo);
printf("Inquiry buffer mismatch:");
for (i = 0; i < AHD_LINUX_DV_INQ_LEN; i++) {
if ((i & 0xF) == 0)
printf("\n ");
printf("0x%x:0x0%x ",
((uint8_t *)targ->inq_data)[i],
targ->dv_buffer[i]);
}
printf("\n");
}
#endif
if (ahd_linux_dv_fallback(ahd, devinfo) != 0) {
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
/*
* Do not count "falling back"
* against our retries.
*/
targ->dv_state_retry = 0;
AHD_SET_DV_STATE(ahd, targ, targ->dv_state);
break;
}
case SS_INQ_REFRESH:
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
case SS_TUR:
case SS_RETRY:
AHD_SET_DV_STATE(ahd, targ, targ->dv_state);
if (ahd_cmd_get_transaction_status(cmd)
== CAM_REQUEUE_REQ) {
targ->dv_state_retry--;
} else if ((status & SSQ_FALLBACK) != 0) {
if (ahd_linux_dv_fallback(ahd, devinfo) != 0) {
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_EXIT);
break;
}
/*
* Do not count "falling back"
* against our retries.
*/
targ->dv_state_retry = 0;
} else if ((status & SS_ERRMASK) == EBUSY)
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_BUSY);
if (targ->dv_state_retry < 10)
break;
/* FALLTHROUGH */
default:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Failed DV inquiry, skipping\n");
}
#endif
break;
}
break;
case AHD_DV_STATE_TUR:
switch (status & SS_MASK) {
case SS_NOP:
if ((targ->flags & AHD_BASIC_DV) != 0) {
ahd_linux_filter_inquiry(ahd, devinfo);
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_VERIFY);
} else if ((targ->flags & AHD_ENHANCED_DV) != 0) {
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_REBD);
} else {
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
}
break;
case SS_RETRY:
case SS_TUR:
if ((status & SS_ERRMASK) == EBUSY) {
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_BUSY);
break;
}
AHD_SET_DV_STATE(ahd, targ, targ->dv_state);
if (ahd_cmd_get_transaction_status(cmd)
== CAM_REQUEUE_REQ) {
targ->dv_state_retry--;
} else if ((status & SSQ_FALLBACK) != 0) {
if (ahd_linux_dv_fallback(ahd, devinfo) != 0) {
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_EXIT);
break;
}
/*
* Do not count "falling back"
* against our retries.
*/
targ->dv_state_retry = 0;
}
if (targ->dv_state_retry >= 10) {
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("DV TUR reties exhausted\n");
}
#endif
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
if (status & SSQ_DELAY)
ssleep(1);
break;
case SS_START:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_SU);
break;
case SS_INQ_REFRESH:
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
default:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
break;
case AHD_DV_STATE_REBD:
switch (status & SS_MASK) {
case SS_NOP:
{
uint32_t echo_size;
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_WEB);
echo_size = scsi_3btoul(&targ->dv_buffer[1]);
echo_size &= 0x1FFF;
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Echo buffer size= %d\n", echo_size);
}
#endif
if (echo_size == 0) {
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
/* Generate the buffer pattern */
targ->dv_echo_size = echo_size;
ahd_linux_generate_dv_pattern(targ);
/*
* Setup initial negotiation values.
*/
ahd_linux_filter_inquiry(ahd, devinfo);
break;
}
case SS_INQ_REFRESH:
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
case SS_RETRY:
AHD_SET_DV_STATE(ahd, targ, targ->dv_state);
if (ahd_cmd_get_transaction_status(cmd)
== CAM_REQUEUE_REQ)
targ->dv_state_retry--;
if (targ->dv_state_retry <= 10)
break;
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("DV REBD reties exhausted\n");
}
#endif
/* FALLTHROUGH */
case SS_FATAL:
default:
/*
* Setup initial negotiation values
* and try level 1 DV.
*/
ahd_linux_filter_inquiry(ahd, devinfo);
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_INQ_VERIFY);
targ->dv_echo_size = 0;
break;
}
break;
case AHD_DV_STATE_WEB:
switch (status & SS_MASK) {
case SS_NOP:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_REB);
break;
case SS_INQ_REFRESH:
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
case SS_RETRY:
AHD_SET_DV_STATE(ahd, targ, targ->dv_state);
if (ahd_cmd_get_transaction_status(cmd)
== CAM_REQUEUE_REQ) {
targ->dv_state_retry--;
} else if ((status & SSQ_FALLBACK) != 0) {
if (ahd_linux_dv_fallback(ahd, devinfo) != 0) {
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_EXIT);
break;
}
/*
* Do not count "falling back"
* against our retries.
*/
targ->dv_state_retry = 0;
}
if (targ->dv_state_retry <= 10)
break;
/* FALLTHROUGH */
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("DV WEB reties exhausted\n");
}
#endif
default:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
break;
case AHD_DV_STATE_REB:
switch (status & SS_MASK) {
case SS_NOP:
if (memcmp(targ->dv_buffer, targ->dv_buffer1,
targ->dv_echo_size) != 0) {
if (ahd_linux_dv_fallback(ahd, devinfo) != 0)
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_EXIT);
else
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_WEB);
break;
}
if (targ->dv_buffer != NULL) {
free(targ->dv_buffer, M_DEVBUF);
targ->dv_buffer = NULL;
}
if (targ->dv_buffer1 != NULL) {
free(targ->dv_buffer1, M_DEVBUF);
targ->dv_buffer1 = NULL;
}
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
case SS_INQ_REFRESH:
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
case SS_RETRY:
AHD_SET_DV_STATE(ahd, targ, targ->dv_state);
if (ahd_cmd_get_transaction_status(cmd)
== CAM_REQUEUE_REQ) {
targ->dv_state_retry--;
} else if ((status & SSQ_FALLBACK) != 0) {
if (ahd_linux_dv_fallback(ahd, devinfo) != 0) {
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_EXIT);
break;
}
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_WEB);
}
if (targ->dv_state_retry <= 10) {
if ((status & (SSQ_DELAY_RANDOM|SSQ_DELAY))!= 0)
msleep(ahd->our_id*1000/10);
break;
}
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("DV REB reties exhausted\n");
}
#endif
/* FALLTHROUGH */
default:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
break;
case AHD_DV_STATE_SU:
switch (status & SS_MASK) {
case SS_NOP:
case SS_INQ_REFRESH:
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
default:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
break;
case AHD_DV_STATE_BUSY:
switch (status & SS_MASK) {
case SS_NOP:
case SS_INQ_REFRESH:
AHD_SET_DV_STATE(ahd, targ,
AHD_DV_STATE_INQ_SHORT_ASYNC);
break;
case SS_TUR:
case SS_RETRY:
AHD_SET_DV_STATE(ahd, targ, targ->dv_state);
if (ahd_cmd_get_transaction_status(cmd)
== CAM_REQUEUE_REQ) {
targ->dv_state_retry--;
} else if (targ->dv_state_retry < 60) {
if ((status & SSQ_DELAY) != 0)
ssleep(1);
} else {
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("DV BUSY reties exhausted\n");
}
#endif
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
}
break;
default:
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
break;
default:
printf("%s: Invalid DV completion state %d\n", ahd_name(ahd),
targ->dv_state);
AHD_SET_DV_STATE(ahd, targ, AHD_DV_STATE_EXIT);
break;
}
}
static void
ahd_linux_dv_fill_cmd(struct ahd_softc *ahd, struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo)
{
memset(cmd, 0, sizeof(struct scsi_cmnd));
cmd->device = ahd->platform_data->dv_scsi_dev;
cmd->scsi_done = ahd_linux_dv_complete;
}
/*
* Synthesize an inquiry command. On the return trip, it'll be
* sniffed and the device transfer settings set for us.
*/
static void
ahd_linux_dv_inq(struct ahd_softc *ahd, struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo, struct ahd_linux_target *targ,
u_int request_length)
{
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Sending INQ\n");
}
#endif
if (targ->inq_data == NULL)
targ->inq_data = malloc(AHD_LINUX_DV_INQ_LEN,
M_DEVBUF, M_WAITOK);
if (targ->dv_state > AHD_DV_STATE_INQ_ASYNC) {
if (targ->dv_buffer != NULL)
free(targ->dv_buffer, M_DEVBUF);
targ->dv_buffer = malloc(AHD_LINUX_DV_INQ_LEN,
M_DEVBUF, M_WAITOK);
}
ahd_linux_dv_fill_cmd(ahd, cmd, devinfo);
cmd->sc_data_direction = DMA_FROM_DEVICE;
cmd->cmd_len = 6;
cmd->cmnd[0] = INQUIRY;
cmd->cmnd[4] = request_length;
cmd->request_bufflen = request_length;
if (targ->dv_state > AHD_DV_STATE_INQ_ASYNC)
cmd->request_buffer = targ->dv_buffer;
else
cmd->request_buffer = targ->inq_data;
memset(cmd->request_buffer, 0, AHD_LINUX_DV_INQ_LEN);
}
static void
ahd_linux_dv_tur(struct ahd_softc *ahd, struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo)
{
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Sending TUR\n");
}
#endif
/* Do a TUR to clear out any non-fatal transitional state */
ahd_linux_dv_fill_cmd(ahd, cmd, devinfo);
cmd->sc_data_direction = DMA_NONE;
cmd->cmd_len = 6;
cmd->cmnd[0] = TEST_UNIT_READY;
}
#define AHD_REBD_LEN 4
static void
ahd_linux_dv_rebd(struct ahd_softc *ahd, struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo, struct ahd_linux_target *targ)
{
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Sending REBD\n");
}
#endif
if (targ->dv_buffer != NULL)
free(targ->dv_buffer, M_DEVBUF);
targ->dv_buffer = malloc(AHD_REBD_LEN, M_DEVBUF, M_WAITOK);
ahd_linux_dv_fill_cmd(ahd, cmd, devinfo);
cmd->sc_data_direction = DMA_FROM_DEVICE;
cmd->cmd_len = 10;
cmd->cmnd[0] = READ_BUFFER;
cmd->cmnd[1] = 0x0b;
scsi_ulto3b(AHD_REBD_LEN, &cmd->cmnd[6]);
cmd->request_bufflen = AHD_REBD_LEN;
cmd->underflow = cmd->request_bufflen;
cmd->request_buffer = targ->dv_buffer;
}
static void
ahd_linux_dv_web(struct ahd_softc *ahd, struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo, struct ahd_linux_target *targ)
{
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Sending WEB\n");
}
#endif
ahd_linux_dv_fill_cmd(ahd, cmd, devinfo);
cmd->sc_data_direction = DMA_TO_DEVICE;
cmd->cmd_len = 10;
cmd->cmnd[0] = WRITE_BUFFER;
cmd->cmnd[1] = 0x0a;
scsi_ulto3b(targ->dv_echo_size, &cmd->cmnd[6]);
cmd->request_bufflen = targ->dv_echo_size;
cmd->underflow = cmd->request_bufflen;
cmd->request_buffer = targ->dv_buffer;
}
static void
ahd_linux_dv_reb(struct ahd_softc *ahd, struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo, struct ahd_linux_target *targ)
{
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Sending REB\n");
}
#endif
ahd_linux_dv_fill_cmd(ahd, cmd, devinfo);
cmd->sc_data_direction = DMA_FROM_DEVICE;
cmd->cmd_len = 10;
cmd->cmnd[0] = READ_BUFFER;
cmd->cmnd[1] = 0x0a;
scsi_ulto3b(targ->dv_echo_size, &cmd->cmnd[6]);
cmd->request_bufflen = targ->dv_echo_size;
cmd->underflow = cmd->request_bufflen;
cmd->request_buffer = targ->dv_buffer1;
}
static void
ahd_linux_dv_su(struct ahd_softc *ahd, struct scsi_cmnd *cmd,
struct ahd_devinfo *devinfo,
struct ahd_linux_target *targ)
{
u_int le;
le = SID_IS_REMOVABLE(targ->inq_data) ? SSS_LOEJ : 0;
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Sending SU\n");
}
#endif
ahd_linux_dv_fill_cmd(ahd, cmd, devinfo);
cmd->sc_data_direction = DMA_NONE;
cmd->cmd_len = 6;
cmd->cmnd[0] = START_STOP_UNIT;
cmd->cmnd[4] = le | SSS_START;
}
static int
ahd_linux_fallback(struct ahd_softc *ahd, struct ahd_devinfo *devinfo)
{
struct ahd_linux_target *targ;
struct ahd_initiator_tinfo *tinfo;
struct ahd_transinfo *goal;
struct ahd_tmode_tstate *tstate;
u_int width;
u_int period;
u_int offset;
u_int ppr_options;
u_int cur_speed;
u_int wide_speed;
u_int narrow_speed;
u_int fallback_speed;
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
ahd_print_devinfo(ahd, devinfo);
printf("Trying to fallback\n");
}
#endif
targ = ahd->platform_data->targets[devinfo->target_offset];
tinfo = ahd_fetch_transinfo(ahd, devinfo->channel,
devinfo->our_scsiid,
devinfo->target, &tstate);
goal = &tinfo->goal;
width = goal->width;
period = goal->period;
offset = goal->offset;
ppr_options = goal->ppr_options;
if (offset == 0)
period = AHD_ASYNC_XFER_PERIOD;
if (targ->dv_next_narrow_period == 0)
targ->dv_next_narrow_period = MAX(period, AHD_SYNCRATE_ULTRA2);
if (targ->dv_next_wide_period == 0)
targ->dv_next_wide_period = period;
if (targ->dv_max_width == 0)
targ->dv_max_width = width;
if (targ->dv_max_ppr_options == 0)
targ->dv_max_ppr_options = ppr_options;
if (targ->dv_last_ppr_options == 0)
targ->dv_last_ppr_options = ppr_options;
cur_speed = aic_calc_speed(width, period, offset, AHD_SYNCRATE_MIN);
wide_speed = aic_calc_speed(MSG_EXT_WDTR_BUS_16_BIT,
targ->dv_next_wide_period,
MAX_OFFSET, AHD_SYNCRATE_MIN);
narrow_speed = aic_calc_speed(MSG_EXT_WDTR_BUS_8_BIT,
targ->dv_next_narrow_period,
MAX_OFFSET, AHD_SYNCRATE_MIN);
fallback_speed = aic_calc_speed(width, period+1, offset,
AHD_SYNCRATE_MIN);
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
printf("cur_speed= %d, wide_speed= %d, narrow_speed= %d, "
"fallback_speed= %d\n", cur_speed, wide_speed,
narrow_speed, fallback_speed);
}
#endif
if (cur_speed > 160000) {
/*
* Paced/DT/IU_REQ only transfer speeds. All we
* can do is fallback in terms of syncrate.
*/
period++;
} else if (cur_speed > 80000) {
if ((ppr_options & MSG_EXT_PPR_IU_REQ) != 0) {
/*
* Try without IU_REQ as it may be confusing
* an expander.
*/
ppr_options &= ~MSG_EXT_PPR_IU_REQ;
} else {
/*
* Paced/DT only transfer speeds. All we
* can do is fallback in terms of syncrate.
*/
period++;
ppr_options = targ->dv_max_ppr_options;
}
} else if (cur_speed > 3300) {
/*
* In this range we the following
* options ordered from highest to
* lowest desireability:
*
* o Wide/DT
* o Wide/non-DT
* o Narrow at a potentally higher sync rate.
*
* All modes are tested with and without IU_REQ
* set since using IUs may confuse an expander.
*/
if ((ppr_options & MSG_EXT_PPR_IU_REQ) != 0) {
ppr_options &= ~MSG_EXT_PPR_IU_REQ;
} else if ((ppr_options & MSG_EXT_PPR_DT_REQ) != 0) {
/*
* Try going non-DT.
*/
ppr_options = targ->dv_max_ppr_options;
ppr_options &= ~MSG_EXT_PPR_DT_REQ;
} else if (targ->dv_last_ppr_options != 0) {
/*
* Try without QAS or any other PPR options.
* We may need a non-PPR message to work with
* an expander. We look at the "last PPR options"
* so we will perform this fallback even if the
* target responded to our PPR negotiation with
* no option bits set.
*/
ppr_options = 0;
} else if (width == MSG_EXT_WDTR_BUS_16_BIT) {
/*
* If the next narrow speed is greater than
* the next wide speed, fallback to narrow.
* Otherwise fallback to the next DT/Wide setting.
* The narrow async speed will always be smaller
* than the wide async speed, so handle this case
* specifically.
*/
ppr_options = targ->dv_max_ppr_options;
if (narrow_speed > fallback_speed
|| period >= AHD_ASYNC_XFER_PERIOD) {
targ->dv_next_wide_period = period+1;
width = MSG_EXT_WDTR_BUS_8_BIT;
period = targ->dv_next_narrow_period;
} else {
period++;
}
} else if ((ahd->features & AHD_WIDE) != 0
&& targ->dv_max_width != 0
&& wide_speed >= fallback_speed
&& (targ->dv_next_wide_period <= AHD_ASYNC_XFER_PERIOD
|| period >= AHD_ASYNC_XFER_PERIOD)) {
/*
* We are narrow. Try falling back
* to the next wide speed with
* all supported ppr options set.
*/
targ->dv_next_narrow_period = period+1;
width = MSG_EXT_WDTR_BUS_16_BIT;
period = targ->dv_next_wide_period;
ppr_options = targ->dv_max_ppr_options;
} else {
/* Only narrow fallback is allowed. */
period++;
ppr_options = targ->dv_max_ppr_options;
}
} else {
return (-1);
}
offset = MAX_OFFSET;
ahd_find_syncrate(ahd, &period, &ppr_options, AHD_SYNCRATE_PACED);
ahd_set_width(ahd, devinfo, width, AHD_TRANS_GOAL, FALSE);
if (period == 0) {
period = 0;
offset = 0;
ppr_options = 0;
if (width == MSG_EXT_WDTR_BUS_8_BIT)
targ->dv_next_narrow_period = AHD_ASYNC_XFER_PERIOD;
else
targ->dv_next_wide_period = AHD_ASYNC_XFER_PERIOD;
}
ahd_set_syncrate(ahd, devinfo, period, offset,
ppr_options, AHD_TRANS_GOAL, FALSE);
targ->dv_last_ppr_options = ppr_options;
return (0);
}
static void
ahd_linux_dv_timeout(struct scsi_cmnd *cmd)
{
struct ahd_softc *ahd;
struct scb *scb;
u_long flags;
ahd = *((struct ahd_softc **)cmd->device->host->hostdata);
ahd_lock(ahd, &flags);
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV) {
printf("%s: Timeout while doing DV command %x.\n",
ahd_name(ahd), cmd->cmnd[0]);
ahd_dump_card_state(ahd);
}
#endif
/*
* Guard against "done race". No action is
* required if we just completed.
*/
if ((scb = (struct scb *)cmd->host_scribble) == NULL) {
ahd_unlock(ahd, &flags);
return;
}
/*
* Command has not completed. Mark this
* SCB as having failing status prior to
* resetting the bus, so we get the correct
* error code.
*/
if ((scb->flags & SCB_SENSE) != 0)
ahd_set_transaction_status(scb, CAM_AUTOSENSE_FAIL);
else
ahd_set_transaction_status(scb, CAM_CMD_TIMEOUT);
ahd_reset_channel(ahd, cmd->device->channel + 'A', /*initiate*/TRUE);
/*
* Add a minimal bus settle delay for devices that are slow to
* respond after bus resets.
*/
ahd_freeze_simq(ahd);
init_timer(&ahd->platform_data->reset_timer);
ahd->platform_data->reset_timer.data = (u_long)ahd;
ahd->platform_data->reset_timer.expires = jiffies + HZ / 2;
ahd->platform_data->reset_timer.function =
(ahd_linux_callback_t *)ahd_release_simq;
add_timer(&ahd->platform_data->reset_timer);
if (ahd_linux_next_device_to_run(ahd) != NULL)
ahd_schedule_runq(ahd);
ahd_linux_run_complete_queue(ahd);
ahd_unlock(ahd, &flags);
}
static void
ahd_linux_dv_complete(struct scsi_cmnd *cmd)
{
struct ahd_softc *ahd;
ahd = *((struct ahd_softc **)cmd->device->host->hostdata);
/* Delete the DV timer before it goes off! */
scsi_delete_timer(cmd);
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_DV)
printf("%s:%c:%d: Command completed, status= 0x%x\n",
ahd_name(ahd), cmd->device->channel, cmd->device->id,
cmd->result);
#endif
/* Wake up the state machine */
up(&ahd->platform_data->dv_cmd_sem);
}
static void
ahd_linux_generate_dv_pattern(struct ahd_linux_target *targ)
{
uint16_t b;
u_int i;
u_int j;
if (targ->dv_buffer != NULL)
free(targ->dv_buffer, M_DEVBUF);
targ->dv_buffer = malloc(targ->dv_echo_size, M_DEVBUF, M_WAITOK);
if (targ->dv_buffer1 != NULL)
free(targ->dv_buffer1, M_DEVBUF);
targ->dv_buffer1 = malloc(targ->dv_echo_size, M_DEVBUF, M_WAITOK);
i = 0;
b = 0x0001;
for (j = 0 ; i < targ->dv_echo_size; j++) {
if (j < 32) {
/*
* 32bytes of sequential numbers.
*/
targ->dv_buffer[i++] = j & 0xff;
} else if (j < 48) {
/*
* 32bytes of repeating 0x0000, 0xffff.
*/
targ->dv_buffer[i++] = (j & 0x02) ? 0xff : 0x00;
} else if (j < 64) {
/*
* 32bytes of repeating 0x5555, 0xaaaa.
*/
targ->dv_buffer[i++] = (j & 0x02) ? 0xaa : 0x55;
} else {
/*
* Remaining buffer is filled with a repeating
* patter of:
*
* 0xffff
* ~0x0001 << shifted once in each loop.
*/
if (j & 0x02) {
if (j & 0x01) {
targ->dv_buffer[i++] = ~(b >> 8) & 0xff;
b <<= 1;
if (b == 0x0000)
b = 0x0001;
} else {
targ->dv_buffer[i++] = (~b & 0xff);
}
} else {
targ->dv_buffer[i++] = 0xff;
}
}
}
}
static u_int
ahd_linux_user_tagdepth(struct ahd_softc *ahd, struct ahd_devinfo *devinfo)
{
static int warned_user;
u_int tags;
tags = 0;
if ((ahd->user_discenable & devinfo->target_mask) != 0) {
if (ahd->unit >= NUM_ELEMENTS(aic79xx_tag_info)) {
if (warned_user == 0) {
printf(KERN_WARNING
"aic79xx: WARNING: Insufficient tag_info instances\n"
"aic79xx: for installed controllers. Using defaults\n"
"aic79xx: Please update the aic79xx_tag_info array in\n"
"aic79xx: the aic79xx_osm.c source file.\n");
warned_user++;
}
tags = AHD_MAX_QUEUE;
} else {
adapter_tag_info_t *tag_info;
tag_info = &aic79xx_tag_info[ahd->unit];
tags = tag_info->tag_commands[devinfo->target_offset];
if (tags > AHD_MAX_QUEUE)
tags = AHD_MAX_QUEUE;
}
}
return (tags);
}
static u_int
ahd_linux_user_dv_setting(struct ahd_softc *ahd)
{
static int warned_user;
int dv;
if (ahd->unit >= NUM_ELEMENTS(aic79xx_dv_settings)) {
if (warned_user == 0) {
printf(KERN_WARNING
"aic79xx: WARNING: Insufficient dv settings instances\n"
"aic79xx: for installed controllers. Using defaults\n"
"aic79xx: Please update the aic79xx_dv_settings array in"
"aic79xx: the aic79xx_osm.c source file.\n");
warned_user++;
}
dv = -1;
} else {
dv = aic79xx_dv_settings[ahd->unit];
}
if (dv < 0) {
/*
* Apply the default.
*/
dv = 1;
if (ahd->seep_config != 0)
dv = (ahd->seep_config->bios_control & CFENABLEDV);
}
return (dv);
}
static void
ahd_linux_setup_user_rd_strm_settings(struct ahd_softc *ahd)
{
static int warned_user;
u_int rd_strm_mask;
u_int target_id;
/*
* If we have specific read streaming info for this controller,
* apply it. Otherwise use the defaults.
*/
if (ahd->unit >= NUM_ELEMENTS(aic79xx_rd_strm_info)) {
if (warned_user == 0) {
printf(KERN_WARNING
"aic79xx: WARNING: Insufficient rd_strm instances\n"
"aic79xx: for installed controllers. Using defaults\n"
"aic79xx: Please update the aic79xx_rd_strm_info array\n"
"aic79xx: in the aic79xx_osm.c source file.\n");
warned_user++;
}
rd_strm_mask = AIC79XX_CONFIGED_RD_STRM;
} else {
rd_strm_mask = aic79xx_rd_strm_info[ahd->unit];
}
for (target_id = 0; target_id < 16; target_id++) {
struct ahd_devinfo devinfo;
struct ahd_initiator_tinfo *tinfo;
struct ahd_tmode_tstate *tstate;
tinfo = ahd_fetch_transinfo(ahd, 'A', ahd->our_id,
target_id, &tstate);
ahd_compile_devinfo(&devinfo, ahd->our_id, target_id,
CAM_LUN_WILDCARD, 'A', ROLE_INITIATOR);
tinfo->user.ppr_options &= ~MSG_EXT_PPR_RD_STRM;
if ((rd_strm_mask & devinfo.target_mask) != 0)
tinfo->user.ppr_options |= MSG_EXT_PPR_RD_STRM;
}
}
/*
* Determines the queue depth for a given device.
*/
static void
ahd_linux_device_queue_depth(struct ahd_softc *ahd,
struct ahd_linux_device *dev)
{
struct ahd_devinfo devinfo;
u_int tags;
ahd_compile_devinfo(&devinfo,
ahd->our_id,
dev->target->target, dev->lun,
dev->target->channel == 0 ? 'A' : 'B',
ROLE_INITIATOR);
tags = ahd_linux_user_tagdepth(ahd, &devinfo);
if (tags != 0
&& dev->scsi_device != NULL
&& dev->scsi_device->tagged_supported != 0) {
ahd_set_tags(ahd, &devinfo, AHD_QUEUE_TAGGED);
ahd_print_devinfo(ahd, &devinfo);
printf("Tagged Queuing enabled. Depth %d\n", tags);
} else {
ahd_set_tags(ahd, &devinfo, AHD_QUEUE_NONE);
}
}
static void
ahd_linux_run_device_queue(struct ahd_softc *ahd, struct ahd_linux_device *dev)
{
struct ahd_cmd *acmd;
struct scsi_cmnd *cmd;
struct scb *scb;
struct hardware_scb *hscb;
struct ahd_initiator_tinfo *tinfo;
struct ahd_tmode_tstate *tstate;
u_int col_idx;
uint16_t mask;
if ((dev->flags & AHD_DEV_ON_RUN_LIST) != 0)
panic("running device on run list");
while ((acmd = TAILQ_FIRST(&dev->busyq)) != NULL
&& dev->openings > 0 && dev->qfrozen == 0) {
/*
* Schedule us to run later. The only reason we are not
* running is because the whole controller Q is frozen.
*/
if (ahd->platform_data->qfrozen != 0
&& AHD_DV_SIMQ_FROZEN(ahd) == 0) {
TAILQ_INSERT_TAIL(&ahd->platform_data->device_runq,
dev, links);
dev->flags |= AHD_DEV_ON_RUN_LIST;
return;
}
cmd = &acmd_scsi_cmd(acmd);
/*
* Get an scb to use.
*/
tinfo = ahd_fetch_transinfo(ahd, 'A', ahd->our_id,
cmd->device->id, &tstate);
if ((dev->flags & (AHD_DEV_Q_TAGGED|AHD_DEV_Q_BASIC)) == 0
|| (tinfo->curr.ppr_options & MSG_EXT_PPR_IU_REQ) != 0) {
col_idx = AHD_NEVER_COL_IDX;
} else {
col_idx = AHD_BUILD_COL_IDX(cmd->device->id,
cmd->device->lun);
}
if ((scb = ahd_get_scb(ahd, col_idx)) == NULL) {
TAILQ_INSERT_TAIL(&ahd->platform_data->device_runq,
dev, links);
dev->flags |= AHD_DEV_ON_RUN_LIST;
ahd->flags |= AHD_RESOURCE_SHORTAGE;
return;
}
TAILQ_REMOVE(&dev->busyq, acmd, acmd_links.tqe);
scb->io_ctx = cmd;
scb->platform_data->dev = dev;
hscb = scb->hscb;
cmd->host_scribble = (char *)scb;
/*
* Fill out basics of the HSCB.
*/
hscb->control = 0;
hscb->scsiid = BUILD_SCSIID(ahd, cmd);
hscb->lun = cmd->device->lun;
scb->hscb->task_management = 0;
mask = SCB_GET_TARGET_MASK(ahd, scb);
if ((ahd->user_discenable & mask) != 0)
hscb->control |= DISCENB;
if (AHD_DV_CMD(cmd) != 0)
scb->flags |= SCB_SILENT;
if ((tinfo->curr.ppr_options & MSG_EXT_PPR_IU_REQ) != 0)
scb->flags |= SCB_PACKETIZED;
if ((tstate->auto_negotiate & mask) != 0) {
scb->flags |= SCB_AUTO_NEGOTIATE;
scb->hscb->control |= MK_MESSAGE;
}
if ((dev->flags & (AHD_DEV_Q_TAGGED|AHD_DEV_Q_BASIC)) != 0) {
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
int msg_bytes;
uint8_t tag_msgs[2];
msg_bytes = scsi_populate_tag_msg(cmd, tag_msgs);
if (msg_bytes && tag_msgs[0] != MSG_SIMPLE_TASK) {
hscb->control |= tag_msgs[0];
if (tag_msgs[0] == MSG_ORDERED_TASK)
dev->commands_since_idle_or_otag = 0;
} else
#endif
if (dev->commands_since_idle_or_otag == AHD_OTAG_THRESH
&& (dev->flags & AHD_DEV_Q_TAGGED) != 0) {
hscb->control |= MSG_ORDERED_TASK;
dev->commands_since_idle_or_otag = 0;
} else {
hscb->control |= MSG_SIMPLE_TASK;
}
}
hscb->cdb_len = cmd->cmd_len;
memcpy(hscb->shared_data.idata.cdb, cmd->cmnd, hscb->cdb_len);
scb->sg_count = 0;
ahd_set_residual(scb, 0);
ahd_set_sense_residual(scb, 0);
if (cmd->use_sg != 0) {
void *sg;
struct scatterlist *cur_seg;
u_int nseg;
int dir;
cur_seg = (struct scatterlist *)cmd->request_buffer;
dir = cmd->sc_data_direction;
nseg = pci_map_sg(ahd->dev_softc, cur_seg,
cmd->use_sg, dir);
scb->platform_data->xfer_len = 0;
for (sg = scb->sg_list; nseg > 0; nseg--, cur_seg++) {
dma_addr_t addr;
bus_size_t len;
addr = sg_dma_address(cur_seg);
len = sg_dma_len(cur_seg);
scb->platform_data->xfer_len += len;
sg = ahd_sg_setup(ahd, scb, sg, addr, len,
/*last*/nseg == 1);
}
} else if (cmd->request_bufflen != 0) {
void *sg;
dma_addr_t addr;
int dir;
sg = scb->sg_list;
dir = cmd->sc_data_direction;
addr = pci_map_single(ahd->dev_softc,
cmd->request_buffer,
cmd->request_bufflen, dir);
scb->platform_data->xfer_len = cmd->request_bufflen;
scb->platform_data->buf_busaddr = addr;
sg = ahd_sg_setup(ahd, scb, sg, addr,
cmd->request_bufflen, /*last*/TRUE);
}
LIST_INSERT_HEAD(&ahd->pending_scbs, scb, pending_links);
dev->openings--;
dev->active++;
dev->commands_issued++;
/* Update the error counting bucket and dump if needed */
if (dev->target->cmds_since_error) {
dev->target->cmds_since_error++;
if (dev->target->cmds_since_error >
AHD_LINUX_ERR_THRESH)
dev->target->cmds_since_error = 0;
}
if ((dev->flags & AHD_DEV_PERIODIC_OTAG) != 0)
dev->commands_since_idle_or_otag++;
scb->flags |= SCB_ACTIVE;
ahd_queue_scb(ahd, scb);
}
}
/*
* SCSI controller interrupt handler.
*/
irqreturn_t
ahd_linux_isr(int irq, void *dev_id, struct pt_regs * regs)
{
struct ahd_softc *ahd;
u_long flags;
int ours;
ahd = (struct ahd_softc *) dev_id;
ahd_lock(ahd, &flags);
ours = ahd_intr(ahd);
if (ahd_linux_next_device_to_run(ahd) != NULL)
ahd_schedule_runq(ahd);
ahd_linux_run_complete_queue(ahd);
ahd_unlock(ahd, &flags);
return IRQ_RETVAL(ours);
}
void
ahd_platform_flushwork(struct ahd_softc *ahd)
{
while (ahd_linux_run_complete_queue(ahd) != NULL)
;
}
static struct ahd_linux_target*
ahd_linux_alloc_target(struct ahd_softc *ahd, u_int channel, u_int target)
{
struct ahd_linux_target *targ;
targ = malloc(sizeof(*targ), M_DEVBUF, M_NOWAIT);
if (targ == NULL)
return (NULL);
memset(targ, 0, sizeof(*targ));
targ->channel = channel;
targ->target = target;
targ->ahd = ahd;
targ->flags = AHD_DV_REQUIRED;
ahd->platform_data->targets[target] = targ;
return (targ);
}
static void
ahd_linux_free_target(struct ahd_softc *ahd, struct ahd_linux_target *targ)
{
struct ahd_devinfo devinfo;
struct ahd_initiator_tinfo *tinfo;
struct ahd_tmode_tstate *tstate;
u_int our_id;
u_int target_offset;
char channel;
/*
* Force a negotiation to async/narrow on any
* future command to this device unless a bus
* reset occurs between now and that command.
*/
channel = 'A' + targ->channel;
our_id = ahd->our_id;
target_offset = targ->target;
tinfo = ahd_fetch_transinfo(ahd, channel, our_id,
targ->target, &tstate);
ahd_compile_devinfo(&devinfo, our_id, targ->target, CAM_LUN_WILDCARD,
channel, ROLE_INITIATOR);
ahd_set_syncrate(ahd, &devinfo, 0, 0, 0,
AHD_TRANS_GOAL, /*paused*/FALSE);
ahd_set_width(ahd, &devinfo, MSG_EXT_WDTR_BUS_8_BIT,
AHD_TRANS_GOAL, /*paused*/FALSE);
ahd_update_neg_request(ahd, &devinfo, tstate, tinfo, AHD_NEG_ALWAYS);
ahd->platform_data->targets[target_offset] = NULL;
if (targ->inq_data != NULL)
free(targ->inq_data, M_DEVBUF);
if (targ->dv_buffer != NULL)
free(targ->dv_buffer, M_DEVBUF);
if (targ->dv_buffer1 != NULL)
free(targ->dv_buffer1, M_DEVBUF);
free(targ, M_DEVBUF);
}
static struct ahd_linux_device*
ahd_linux_alloc_device(struct ahd_softc *ahd,
struct ahd_linux_target *targ, u_int lun)
{
struct ahd_linux_device *dev;
dev = malloc(sizeof(*dev), M_DEVBUG, M_NOWAIT);
if (dev == NULL)
return (NULL);
memset(dev, 0, sizeof(*dev));
init_timer(&dev->timer);
TAILQ_INIT(&dev->busyq);
dev->flags = AHD_DEV_UNCONFIGURED;
dev->lun = lun;
dev->target = targ;
/*
* We start out life using untagged
* transactions of which we allow one.
*/
dev->openings = 1;
/*
* Set maxtags to 0. This will be changed if we
* later determine that we are dealing with
* a tagged queuing capable device.
*/
dev->maxtags = 0;
targ->refcount++;
targ->devices[lun] = dev;
return (dev);
}
static void
ahd_linux_free_device(struct ahd_softc *ahd, struct ahd_linux_device *dev)
{
struct ahd_linux_target *targ;
del_timer(&dev->timer);
targ = dev->target;
targ->devices[dev->lun] = NULL;
free(dev, M_DEVBUF);
targ->refcount--;
if (targ->refcount == 0
&& (targ->flags & AHD_DV_REQUIRED) == 0)
ahd_linux_free_target(ahd, targ);
}
void
ahd_send_async(struct ahd_softc *ahd, char channel,
u_int target, u_int lun, ac_code code, void *arg)
{
switch (code) {
case AC_TRANSFER_NEG:
{
char buf[80];
struct ahd_linux_target *targ;
struct info_str info;
struct ahd_initiator_tinfo *tinfo;
struct ahd_tmode_tstate *tstate;
info.buffer = buf;
info.length = sizeof(buf);
info.offset = 0;
info.pos = 0;
tinfo = ahd_fetch_transinfo(ahd, channel, ahd->our_id,
target, &tstate);
/*
* Don't bother reporting results while
* negotiations are still pending.
*/
if (tinfo->curr.period != tinfo->goal.period
|| tinfo->curr.width != tinfo->goal.width
|| tinfo->curr.offset != tinfo->goal.offset
|| tinfo->curr.ppr_options != tinfo->goal.ppr_options)
if (bootverbose == 0)
break;
/*
* Don't bother reporting results that
* are identical to those last reported.
*/
targ = ahd->platform_data->targets[target];
if (targ == NULL)
break;
if (tinfo->curr.period == targ->last_tinfo.period
&& tinfo->curr.width == targ->last_tinfo.width
&& tinfo->curr.offset == targ->last_tinfo.offset
&& tinfo->curr.ppr_options == targ->last_tinfo.ppr_options)
if (bootverbose == 0)
break;
targ->last_tinfo.period = tinfo->curr.period;
targ->last_tinfo.width = tinfo->curr.width;
targ->last_tinfo.offset = tinfo->curr.offset;
targ->last_tinfo.ppr_options = tinfo->curr.ppr_options;
printf("(%s:%c:", ahd_name(ahd), channel);
if (target == CAM_TARGET_WILDCARD)
printf("*): ");
else
printf("%d): ", target);
ahd_format_transinfo(&info, &tinfo->curr);
if (info.pos < info.length)
*info.buffer = '\0';
else
buf[info.length - 1] = '\0';
printf("%s", buf);
break;
}
case AC_SENT_BDR:
{
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
WARN_ON(lun != CAM_LUN_WILDCARD);
scsi_report_device_reset(ahd->platform_data->host,
channel - 'A', target);
#else
Scsi_Device *scsi_dev;
/*
* Find the SCSI device associated with this
* request and indicate that a UA is expected.
*/
for (scsi_dev = ahd->platform_data->host->host_queue;
scsi_dev != NULL; scsi_dev = scsi_dev->next) {
if (channel - 'A' == scsi_dev->channel
&& target == scsi_dev->id
&& (lun == CAM_LUN_WILDCARD
|| lun == scsi_dev->lun)) {
scsi_dev->was_reset = 1;
scsi_dev->expecting_cc_ua = 1;
}
}
#endif
break;
}
case AC_BUS_RESET:
if (ahd->platform_data->host != NULL) {
scsi_report_bus_reset(ahd->platform_data->host,
channel - 'A');
}
break;
default:
panic("ahd_send_async: Unexpected async event");
}
}
/*
* Calls the higher level scsi done function and frees the scb.
*/
void
ahd_done(struct ahd_softc *ahd, struct scb *scb)
{
Scsi_Cmnd *cmd;
struct ahd_linux_device *dev;
if ((scb->flags & SCB_ACTIVE) == 0) {
printf("SCB %d done'd twice\n", SCB_GET_TAG(scb));
ahd_dump_card_state(ahd);
panic("Stopping for safety");
}
LIST_REMOVE(scb, pending_links);
cmd = scb->io_ctx;
dev = scb->platform_data->dev;
dev->active--;
dev->openings++;
if ((cmd->result & (CAM_DEV_QFRZN << 16)) != 0) {
cmd->result &= ~(CAM_DEV_QFRZN << 16);
dev->qfrozen--;
}
ahd_linux_unmap_scb(ahd, scb);
/*
* Guard against stale sense data.
* The Linux mid-layer assumes that sense
* was retrieved anytime the first byte of
* the sense buffer looks "sane".
*/
cmd->sense_buffer[0] = 0;
if (ahd_get_transaction_status(scb) == CAM_REQ_INPROG) {
uint32_t amount_xferred;
amount_xferred =
ahd_get_transfer_length(scb) - ahd_get_residual(scb);
if ((scb->flags & SCB_TRANSMISSION_ERROR) != 0) {
#ifdef AHD_DEBUG
if ((ahd_debug & AHD_SHOW_MISC) != 0) {
ahd_print_path(ahd, scb);
printf("Set CAM_UNCOR_PARITY\n");
}
#endif
ahd_set_transaction_status(scb, CAM_UNCOR_PARITY);
#ifdef AHD_REPORT_UNDERFLOWS
/*
* This code is disabled by default as some
* clients of the SCSI system do not properly
* initialize the underflow parameter. This
* results in spurious termination of commands
* that complete as expected (e.g. underflow is
* allowed as command can return variable amounts
* of data.
*/
} else if (amount_xferred < scb->io_ctx->underflow) {
u_int i;
ahd_print_path(ahd, scb);
printf("CDB:");
for (i = 0; i < scb->io_ctx->cmd_len; i++)
printf(" 0x%x", scb->io_ctx->cmnd[i]);
printf("\n");
ahd_print_path(ahd, scb);
printf("Saw underflow (%ld of %ld bytes). "
"Treated as error\n",
ahd_get_residual(scb),
ahd_get_transfer_length(scb));
ahd_set_transaction_status(scb, CAM_DATA_RUN_ERR);
#endif
} else {
ahd_set_transaction_status(scb, CAM_REQ_CMP);
}
} else if (ahd_get_transaction_status(scb) == CAM_SCSI_STATUS_ERROR) {
ahd_linux_handle_scsi_status(ahd, dev, scb);
} else if (ahd_get_transaction_status(scb) == CAM_SEL_TIMEOUT) {
dev->flags |= AHD_DEV_UNCONFIGURED;
if (AHD_DV_CMD(cmd) == FALSE)
dev->target->flags &= ~AHD_DV_REQUIRED;
}
/*
* Start DV for devices that require it assuming the first command
* sent does not result in a selection timeout.
*/
if (ahd_get_transaction_status(scb) != CAM_SEL_TIMEOUT
&& (dev->target->flags & AHD_DV_REQUIRED) != 0)
ahd_linux_start_dv(ahd);
if (dev->openings == 1
&& ahd_get_transaction_status(scb) == CAM_REQ_CMP
&& ahd_get_scsi_status(scb) != SCSI_STATUS_QUEUE_FULL)
dev->tag_success_count++;
/*
* Some devices deal with temporary internal resource
* shortages by returning queue full. When the queue
* full occurrs, we throttle back. Slowly try to get
* back to our previous queue depth.
*/
if ((dev->openings + dev->active) < dev->maxtags
&& dev->tag_success_count > AHD_TAG_SUCCESS_INTERVAL) {
dev->tag_success_count = 0;
dev->openings++;
}
if (dev->active == 0)
dev->commands_since_idle_or_otag = 0;
if (TAILQ_EMPTY(&dev->busyq)) {
if ((dev->flags & AHD_DEV_UNCONFIGURED) != 0
&& dev->active == 0
&& (dev->flags & AHD_DEV_TIMER_ACTIVE) == 0)
ahd_linux_free_device(ahd, dev);
} else if ((dev->flags & AHD_DEV_ON_RUN_LIST) == 0) {
TAILQ_INSERT_TAIL(&ahd->platform_data->device_runq, dev, links);
dev->flags |= AHD_DEV_ON_RUN_LIST;
}
if ((scb->flags & SCB_RECOVERY_SCB) != 0) {
printf("Recovery SCB completes\n");
if (ahd_get_transaction_status(scb) == CAM_BDR_SENT
|| ahd_get_transaction_status(scb) == CAM_REQ_ABORTED)
ahd_set_transaction_status(scb, CAM_CMD_TIMEOUT);
if ((scb->platform_data->flags & AHD_SCB_UP_EH_SEM) != 0) {
scb->platform_data->flags &= ~AHD_SCB_UP_EH_SEM;
up(&ahd->platform_data->eh_sem);
}
}
ahd_free_scb(ahd, scb);
ahd_linux_queue_cmd_complete(ahd, cmd);
if ((ahd->platform_data->flags & AHD_DV_WAIT_SIMQ_EMPTY) != 0
&& LIST_FIRST(&ahd->pending_scbs) == NULL) {
ahd->platform_data->flags &= ~AHD_DV_WAIT_SIMQ_EMPTY;
up(&ahd->platform_data->dv_sem);
}
}
static void
ahd_linux_handle_scsi_status(struct ahd_softc *ahd,
struct ahd_linux_device *dev, struct scb *scb)
{
struct ahd_devinfo devinfo;
ahd_compile_devinfo(&devinfo,
ahd->our_id,
dev->target->target, dev->lun,
dev->target->channel == 0 ? 'A' : 'B',
ROLE_INITIATOR);
/*
* We don't currently trust the mid-layer to
* properly deal with queue full or busy. So,
* when one occurs, we tell the mid-layer to
* unconditionally requeue the command to us
* so that we can retry it ourselves. We also
* implement our own throttling mechanism so
* we don't clobber the device with too many
* commands.
*/
switch (ahd_get_scsi_status(scb)) {
default:
break;
case SCSI_STATUS_CHECK_COND:
case SCSI_STATUS_CMD_TERMINATED:
{
Scsi_Cmnd *cmd;
/*
* Copy sense information to the OS's cmd
* structure if it is available.
*/
cmd = scb->io_ctx;
if ((scb->flags & (SCB_SENSE|SCB_PKT_SENSE)) != 0) {
struct scsi_status_iu_header *siu;
u_int sense_size;
u_int sense_offset;
if (scb->flags & SCB_SENSE) {
sense_size = MIN(sizeof(struct scsi_sense_data)
- ahd_get_sense_residual(scb),
sizeof(cmd->sense_buffer));
sense_offset = 0;
} else {
/*
* Copy only the sense data into the provided
* buffer.
*/
siu = (struct scsi_status_iu_header *)
scb->sense_data;
sense_size = MIN(scsi_4btoul(siu->sense_length),
sizeof(cmd->sense_buffer));
sense_offset = SIU_SENSE_OFFSET(siu);
}
memset(cmd->sense_buffer, 0, sizeof(cmd->sense_buffer));
memcpy(cmd->sense_buffer,
ahd_get_sense_buf(ahd, scb)
+ sense_offset, sense_size);
cmd->result |= (DRIVER_SENSE << 24);
#ifdef AHD_DEBUG
if (ahd_debug & AHD_SHOW_SENSE) {
int i;
printf("Copied %d bytes of sense data at %d:",
sense_size, sense_offset);
for (i = 0; i < sense_size; i++) {
if ((i & 0xF) == 0)
printf("\n");
printf("0x%x ", cmd->sense_buffer[i]);
}
printf("\n");
}
#endif
}
break;
}
case SCSI_STATUS_QUEUE_FULL:
{
/*
* By the time the core driver has returned this
* command, all other commands that were queued
* to us but not the device have been returned.
* This ensures that dev->active is equal to
* the number of commands actually queued to
* the device.
*/
dev->tag_success_count = 0;
if (dev->active != 0) {
/*
* Drop our opening count to the number
* of commands currently outstanding.
*/
dev->openings = 0;
#ifdef AHD_DEBUG
if ((ahd_debug & AHD_SHOW_QFULL) != 0) {
ahd_print_path(ahd, scb);
printf("Dropping tag count to %d\n",
dev->active);
}
#endif
if (dev->active == dev->tags_on_last_queuefull) {
dev->last_queuefull_same_count++;
/*
* If we repeatedly see a queue full
* at the same queue depth, this
* device has a fixed number of tag
* slots. Lock in this tag depth
* so we stop seeing queue fulls from
* this device.
*/
if (dev->last_queuefull_same_count
== AHD_LOCK_TAGS_COUNT) {
dev->maxtags = dev->active;
ahd_print_path(ahd, scb);
printf("Locking max tag count at %d\n",
dev->active);
}
} else {
dev->tags_on_last_queuefull = dev->active;
dev->last_queuefull_same_count = 0;
}
ahd_set_transaction_status(scb, CAM_REQUEUE_REQ);
ahd_set_scsi_status(scb, SCSI_STATUS_OK);
ahd_platform_set_tags(ahd, &devinfo,
(dev->flags & AHD_DEV_Q_BASIC)
? AHD_QUEUE_BASIC : AHD_QUEUE_TAGGED);
break;
}
/*
* Drop down to a single opening, and treat this
* as if the target returned BUSY SCSI status.
*/
dev->openings = 1;
ahd_platform_set_tags(ahd, &devinfo,
(dev->flags & AHD_DEV_Q_BASIC)
? AHD_QUEUE_BASIC : AHD_QUEUE_TAGGED);
ahd_set_scsi_status(scb, SCSI_STATUS_BUSY);
/* FALLTHROUGH */
}
case SCSI_STATUS_BUSY:
/*
* Set a short timer to defer sending commands for
* a bit since Linux will not delay in this case.
*/
if ((dev->flags & AHD_DEV_TIMER_ACTIVE) != 0) {
printf("%s:%c:%d: Device Timer still active during "
"busy processing\n", ahd_name(ahd),
dev->target->channel, dev->target->target);
break;
}
dev->flags |= AHD_DEV_TIMER_ACTIVE;
dev->qfrozen++;
init_timer(&dev->timer);
dev->timer.data = (u_long)dev;
dev->timer.expires = jiffies + (HZ/2);
dev->timer.function = ahd_linux_dev_timed_unfreeze;
add_timer(&dev->timer);
break;
}
}
static void
ahd_linux_queue_cmd_complete(struct ahd_softc *ahd, Scsi_Cmnd *cmd)
{
/*
* Typically, the complete queue has very few entries
* queued to it before the queue is emptied by
* ahd_linux_run_complete_queue, so sorting the entries
* by generation number should be inexpensive.
* We perform the sort so that commands that complete
* with an error are retuned in the order origionally
* queued to the controller so that any subsequent retries
* are performed in order. The underlying ahd routines do
* not guarantee the order that aborted commands will be
* returned to us.
*/
struct ahd_completeq *completeq;
struct ahd_cmd *list_cmd;
struct ahd_cmd *acmd;
/*
* Map CAM error codes into Linux Error codes. We
* avoid the conversion so that the DV code has the
* full error information available when making
* state change decisions.
*/
if (AHD_DV_CMD(cmd) == FALSE) {
uint32_t status;
u_int new_status;
status = ahd_cmd_get_transaction_status(cmd);
if (status != CAM_REQ_CMP) {
struct ahd_linux_device *dev;
struct ahd_devinfo devinfo;
cam_status cam_status;
uint32_t action;
u_int scsi_status;
dev = ahd_linux_get_device(ahd, cmd->device->channel,
cmd->device->id,
cmd->device->lun,
/*alloc*/FALSE);
if (dev == NULL)
goto no_fallback;
ahd_compile_devinfo(&devinfo,
ahd->our_id,
dev->target->target, dev->lun,
dev->target->channel == 0 ? 'A':'B',
ROLE_INITIATOR);
scsi_status = ahd_cmd_get_scsi_status(cmd);
cam_status = ahd_cmd_get_transaction_status(cmd);
action = aic_error_action(cmd, dev->target->inq_data,
cam_status, scsi_status);
if ((action & SSQ_FALLBACK) != 0) {
/* Update stats */
dev->target->errors_detected++;
if (dev->target->cmds_since_error == 0)
dev->target->cmds_since_error++;
else {
dev->target->cmds_since_error = 0;
ahd_linux_fallback(ahd, &devinfo);
}
}
}
no_fallback:
switch (status) {
case CAM_REQ_INPROG:
case CAM_REQ_CMP:
case CAM_SCSI_STATUS_ERROR:
new_status = DID_OK;
break;
case CAM_REQ_ABORTED:
new_status = DID_ABORT;
break;
case CAM_BUSY:
new_status = DID_BUS_BUSY;
break;
case CAM_REQ_INVALID:
case CAM_PATH_INVALID:
new_status = DID_BAD_TARGET;
break;
case CAM_SEL_TIMEOUT:
new_status = DID_NO_CONNECT;
break;
case CAM_SCSI_BUS_RESET:
case CAM_BDR_SENT:
new_status = DID_RESET;
break;
case CAM_UNCOR_PARITY:
new_status = DID_PARITY;
break;
case CAM_CMD_TIMEOUT:
new_status = DID_TIME_OUT;
break;
case CAM_UA_ABORT:
case CAM_REQ_CMP_ERR:
case CAM_AUTOSENSE_FAIL:
case CAM_NO_HBA:
case CAM_DATA_RUN_ERR:
case CAM_UNEXP_BUSFREE:
case CAM_SEQUENCE_FAIL:
case CAM_CCB_LEN_ERR:
case CAM_PROVIDE_FAIL:
case CAM_REQ_TERMIO:
case CAM_UNREC_HBA_ERROR:
case CAM_REQ_TOO_BIG:
new_status = DID_ERROR;
break;
case CAM_REQUEUE_REQ:
/*
* If we want the request requeued, make sure there
* are sufficent retries. In the old scsi error code,
* we used to be able to specify a result code that
* bypassed the retry count. Now we must use this
* hack. We also "fake" a check condition with
* a sense code of ABORTED COMMAND. This seems to
* evoke a retry even if this command is being sent
* via the eh thread. Ick! Ick! Ick!
*/
if (cmd->retries > 0)
cmd->retries--;
new_status = DID_OK;
ahd_cmd_set_scsi_status(cmd, SCSI_STATUS_CHECK_COND);
cmd->result |= (DRIVER_SENSE << 24);
memset(cmd->sense_buffer, 0,
sizeof(cmd->sense_buffer));
cmd->sense_buffer[0] = SSD_ERRCODE_VALID
| SSD_CURRENT_ERROR;
cmd->sense_buffer[2] = SSD_KEY_ABORTED_COMMAND;
break;
default:
/* We should never get here */
new_status = DID_ERROR;
break;
}
ahd_cmd_set_transaction_status(cmd, new_status);
}
completeq = &ahd->platform_data->completeq;
list_cmd = TAILQ_FIRST(completeq);
acmd = (struct ahd_cmd *)cmd;
while (list_cmd != NULL
&& acmd_scsi_cmd(list_cmd).serial_number
< acmd_scsi_cmd(acmd).serial_number)
list_cmd = TAILQ_NEXT(list_cmd, acmd_links.tqe);
if (list_cmd != NULL)
TAILQ_INSERT_BEFORE(list_cmd, acmd, acmd_links.tqe);
else
TAILQ_INSERT_TAIL(completeq, acmd, acmd_links.tqe);
}
static void
ahd_linux_filter_inquiry(struct ahd_softc *ahd, struct ahd_devinfo *devinfo)
{
struct scsi_inquiry_data *sid;
struct ahd_initiator_tinfo *tinfo;
struct ahd_transinfo *user;
struct ahd_transinfo *goal;
struct ahd_transinfo *curr;
struct ahd_tmode_tstate *tstate;
struct ahd_linux_device *dev;
u_int width;
u_int period;
u_int offset;
u_int ppr_options;
u_int trans_version;
u_int prot_version;
/*
* Determine if this lun actually exists. If so,
* hold on to its corresponding device structure.
* If not, make sure we release the device and
* don't bother processing the rest of this inquiry
* command.
*/
dev = ahd_linux_get_device(ahd, devinfo->channel - 'A',
devinfo->target, devinfo->lun,
/*alloc*/TRUE);
sid = (struct scsi_inquiry_data *)dev->target->inq_data;
if (SID_QUAL(sid) == SID_QUAL_LU_CONNECTED) {
dev->flags &= ~AHD_DEV_UNCONFIGURED;
} else {
dev->flags |= AHD_DEV_UNCONFIGURED;
return;
}
/*
* Update our notion of this device's transfer
* negotiation capabilities.
*/
tinfo = ahd_fetch_transinfo(ahd, devinfo->channel,
devinfo->our_scsiid,
devinfo->target, &tstate);
user = &tinfo->user;
goal = &tinfo->goal;
curr = &tinfo->curr;
width = user->width;
period = user->period;
offset = user->offset;
ppr_options = user->ppr_options;
trans_version = user->transport_version;
prot_version = MIN(user->protocol_version, SID_ANSI_REV(sid));
/*
* Only attempt SPI3/4 once we've verified that
* the device claims to support SPI3/4 features.
*/
if (prot_version < SCSI_REV_2)
trans_version = SID_ANSI_REV(sid);
else
trans_version = SCSI_REV_2;
if ((sid->flags & SID_WBus16) == 0)
width = MSG_EXT_WDTR_BUS_8_BIT;
if ((sid->flags & SID_Sync) == 0) {
period = 0;
offset = 0;
ppr_options = 0;
}
if ((sid->spi3data & SID_SPI_QAS) == 0)
ppr_options &= ~MSG_EXT_PPR_QAS_REQ;
if ((sid->spi3data & SID_SPI_CLOCK_DT) == 0)
ppr_options &= MSG_EXT_PPR_QAS_REQ;
if ((sid->spi3data & SID_SPI_IUS) == 0)
ppr_options &= (MSG_EXT_PPR_DT_REQ
| MSG_EXT_PPR_QAS_REQ);
if (prot_version > SCSI_REV_2
&& ppr_options != 0)
trans_version = user->transport_version;
ahd_validate_width(ahd, /*tinfo limit*/NULL, &width, ROLE_UNKNOWN);
ahd_find_syncrate(ahd, &period, &ppr_options, AHD_SYNCRATE_MAX);
ahd_validate_offset(ahd, /*tinfo limit*/NULL, period,
&offset, width, ROLE_UNKNOWN);
if (offset == 0 || period == 0) {
period = 0;
offset = 0;
ppr_options = 0;
}
/* Apply our filtered user settings. */
curr->transport_version = trans_version;
curr->protocol_version = prot_version;
ahd_set_width(ahd, devinfo, width, AHD_TRANS_GOAL, /*paused*/FALSE);
ahd_set_syncrate(ahd, devinfo, period, offset, ppr_options,
AHD_TRANS_GOAL, /*paused*/FALSE);
}
void
ahd_freeze_simq(struct ahd_softc *ahd)
{
ahd->platform_data->qfrozen++;
if (ahd->platform_data->qfrozen == 1) {
scsi_block_requests(ahd->platform_data->host);
ahd_platform_abort_scbs(ahd, CAM_TARGET_WILDCARD, ALL_CHANNELS,
CAM_LUN_WILDCARD, SCB_LIST_NULL,
ROLE_INITIATOR, CAM_REQUEUE_REQ);
}
}
void
ahd_release_simq(struct ahd_softc *ahd)
{
u_long s;
int unblock_reqs;
unblock_reqs = 0;
ahd_lock(ahd, &s);
if (ahd->platform_data->qfrozen > 0)
ahd->platform_data->qfrozen--;
if (ahd->platform_data->qfrozen == 0) {
unblock_reqs = 1;
}
if (AHD_DV_SIMQ_FROZEN(ahd)
&& ((ahd->platform_data->flags & AHD_DV_WAIT_SIMQ_RELEASE) != 0)) {
ahd->platform_data->flags &= ~AHD_DV_WAIT_SIMQ_RELEASE;
up(&ahd->platform_data->dv_sem);
}
ahd_schedule_runq(ahd);
ahd_unlock(ahd, &s);
/*
* There is still a race here. The mid-layer
* should keep its own freeze count and use
* a bottom half handler to run the queues
* so we can unblock with our own lock held.
*/
if (unblock_reqs)
scsi_unblock_requests(ahd->platform_data->host);
}
static void
ahd_linux_sem_timeout(u_long arg)
{
struct scb *scb;
struct ahd_softc *ahd;
u_long s;
scb = (struct scb *)arg;
ahd = scb->ahd_softc;
ahd_lock(ahd, &s);
if ((scb->platform_data->flags & AHD_SCB_UP_EH_SEM) != 0) {
scb->platform_data->flags &= ~AHD_SCB_UP_EH_SEM;
up(&ahd->platform_data->eh_sem);
}
ahd_unlock(ahd, &s);
}
static void
ahd_linux_dev_timed_unfreeze(u_long arg)
{
struct ahd_linux_device *dev;
struct ahd_softc *ahd;
u_long s;
dev = (struct ahd_linux_device *)arg;
ahd = dev->target->ahd;
ahd_lock(ahd, &s);
dev->flags &= ~AHD_DEV_TIMER_ACTIVE;
if (dev->qfrozen > 0)
dev->qfrozen--;
if (dev->qfrozen == 0
&& (dev->flags & AHD_DEV_ON_RUN_LIST) == 0)
ahd_linux_run_device_queue(ahd, dev);
if ((dev->flags & AHD_DEV_UNCONFIGURED) != 0
&& dev->active == 0)
ahd_linux_free_device(ahd, dev);
ahd_unlock(ahd, &s);
}
void
ahd_platform_dump_card_state(struct ahd_softc *ahd)
{
struct ahd_linux_device *dev;
int target;
int maxtarget;
int lun;
int i;
maxtarget = (ahd->features & AHD_WIDE) ? 15 : 7;
for (target = 0; target <=maxtarget; target++) {
for (lun = 0; lun < AHD_NUM_LUNS; lun++) {
struct ahd_cmd *acmd;
dev = ahd_linux_get_device(ahd, 0, target,
lun, /*alloc*/FALSE);
if (dev == NULL)
continue;
printf("DevQ(%d:%d:%d): ", 0, target, lun);
i = 0;
TAILQ_FOREACH(acmd, &dev->busyq, acmd_links.tqe) {
if (i++ > AHD_SCB_MAX)
break;
}
printf("%d waiting\n", i);
}
}
}
static int __init
ahd_linux_init(void)
{
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)
return ahd_linux_detect(&aic79xx_driver_template);
#else
scsi_register_module(MODULE_SCSI_HA, &aic79xx_driver_template);
if (aic79xx_driver_template.present == 0) {
scsi_unregister_module(MODULE_SCSI_HA,
&aic79xx_driver_template);
return (-ENODEV);
}
return (0);
#endif
}
static void __exit
ahd_linux_exit(void)
{
struct ahd_softc *ahd;
/*
* Shutdown DV threads before going into the SCSI mid-layer.
* This avoids situations where the mid-layer locks the entire
* kernel so that waiting for our DV threads to exit leads
* to deadlock.
*/
TAILQ_FOREACH(ahd, &ahd_tailq, links) {
ahd_linux_kill_dv_thread(ahd);
}
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)
/*
* In 2.4 we have to unregister from the PCI core _after_
* unregistering from the scsi midlayer to avoid dangling
* references.
*/
scsi_unregister_module(MODULE_SCSI_HA, &aic79xx_driver_template);
#endif
ahd_linux_pci_exit();
}
module_init(ahd_linux_init);
module_exit(ahd_linux_exit);
Reference in a new issue View git blame Copy permalink