freebsd-dev/sys/dev/amr/amrreg.h

651 lines
23 KiB
C
Raw Normal View History

/*-
* Copyright (c) 1999,2000 Michael Smith
* Copyright (c) 2000 BSDi
* 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.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, 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 DAMAGE.
*
* Copyright (c) 2002 Eric Moore
* Copyright (c) 2002 LSI Logic Corporation
* 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.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The party using or redistributing the source code and binary forms
* agrees to the disclaimer below and the terms and conditions set forth
* herein.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, 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 DAMAGE.
*
*
* $FreeBSD$
*/
/********************************************************************************
********************************************************************************
Driver parameters
********************************************************************************
********************************************************************************/
/*
* We could actually use all 17 segments, but using only 16 means that
* each scatter/gather map is 128 bytes in size, and thus we don't have to worry about
* maps crossing page boundaries.
*
* The AMI documentation says that the limit is 26. Unfortunately, there's no way to
* cleanly fit more than 16 entries in without a page boundary. But is this a concern,
* since we allocate the s/g maps contiguously anyway?
*/
/*
* emoore - Oct 21, 2002
* firmware doesn't have sglist boundary restrictions.
* The sgelem can be set to 26
*/
#define AMR_NSEG 26
#define AMR_MAXCMD 255 /* ident = 0 not allowed */
#define AMR_LIMITCMD 120 /* maximum count of outstanding commands */
#define AMR_MAXLD 40
#define AMR_MAX_CHANNELS 8
#define AMR_MAX_TARGETS 15
#define AMR_MAX_LUNS 7
#define AMR_MAX_SCSI_CMDS (15 * AMR_MAX_CHANNELS) /* one for every target? */
#define AMR_MAX_CDB_LEN 0x0a
#define AMR_MAX_EXTCDB_LEN 0x10
#define AMR_MAX_REQ_SENSE_LEN 0x20
#define AMR_BLKSIZE 512 /* constant for all controllers */
/*
* Perform at-startup board initialisation.
* At this point in time, this code doesn't work correctly, so leave it disabled.
*/
/*#define AMR_BOARD_INIT*/
/********************************************************************************
********************************************************************************
Interface Magic Numbers
********************************************************************************
********************************************************************************/
/*
* Mailbox commands
*/
#define AMR_CMD_LREAD 0x01
#define AMR_CMD_LWRITE 0x02
#define AMR_CMD_PASS 0x03
#define AMR_CMD_EXT_ENQUIRY 0x04
#define AMR_CMD_ENQUIRY 0x05
#define AMR_CMD_FLUSH 0x0a
#define AMR_CMD_EXT_ENQUIRY2 0x0c
#define AMR_CONFIG_PRODINFO 0x0e
#define AMR_CMD_GET_MACHINEID 0x36
#define AMR_CMD_GET_INITIATOR 0x7d /* returns one byte */
#define AMR_CMD_CONFIG 0xa1
Mega update to the LSI MegaRAID driver: 1. Implement a large set of ioctl shims so that the Linux management apps from LSI will work. This includes infrastructure to support adding, deleting and rescanning arrays at runtime. This is based on work from Doug Ambrosko, heavily augmented by LSI and Yahoo. 2. Implement full 64-bit DMA support. Systems with more than 4GB of RAM can now operate without the cost of bounce buffers. Cards that cannot do 64-bit DMA will automatically revert to using bounce buffers. This option can be forced off by setting the 'hw.amr.force_sg32" tunable in the loader. It should only be turned off for debugging purposes. This work was sponsored by Yahoo. 3. Streamline the command delivery and interrupt handler paths after much discussion with Dell and LSI. The logic now closely matches the intended design, making it both more robust and much faster. Certain i/o failures under heavy load should be fixed with this. 4. Optimize the locking. In the interrupt handler, the card can be checked for completed commands without any locks held, due to the handler being implicitely serialized and there being no need to look at any shared data. Only grab the lock to return the command structure to the free pool. A small optimization can still be made to collect all of the completions together and then free them together under a single lock. Items 3 and 4 significantly increase the performance of the driver. On an LSI 320-2X card, transactions per second went from 13,000 to 31,000 in my testing with these changes. However, these changes are still fairly experimental and shouldn't be merged to 6.x until there is more testing. Thanks to Doug Ambrosko, LSI, Dell, and Yahoo for contributing towards this.
2005-12-14 03:26:49 +00:00
#define AMR_CMD_LREAD64 0xa7
#define AMR_CMD_LWRITE64 0xa8
#define AMR_CMD_PASS_64 0xc3
#define AMR_CMD_EXTPASS 0xe3
#define AMR_CONFIG_READ_NVRAM_CONFIG 0x04
#define AMR_CONFIG_WRITE_NVRAM_CONFIG 0x0d
#define AMR_CONFIG_PRODUCT_INFO 0x0e
#define AMR_CONFIG_ENQ3 0x0f
#define AMR_CONFIG_ENQ3_SOLICITED_NOTIFY 0x01
#define AMR_CONFIG_ENQ3_SOLICITED_FULL 0x02
#define AMR_CONFIG_ENQ3_UNSOLICITED 0x03
Mega update to the LSI MegaRAID driver: 1. Implement a large set of ioctl shims so that the Linux management apps from LSI will work. This includes infrastructure to support adding, deleting and rescanning arrays at runtime. This is based on work from Doug Ambrosko, heavily augmented by LSI and Yahoo. 2. Implement full 64-bit DMA support. Systems with more than 4GB of RAM can now operate without the cost of bounce buffers. Cards that cannot do 64-bit DMA will automatically revert to using bounce buffers. This option can be forced off by setting the 'hw.amr.force_sg32" tunable in the loader. It should only be turned off for debugging purposes. This work was sponsored by Yahoo. 3. Streamline the command delivery and interrupt handler paths after much discussion with Dell and LSI. The logic now closely matches the intended design, making it both more robust and much faster. Certain i/o failures under heavy load should be fixed with this. 4. Optimize the locking. In the interrupt handler, the card can be checked for completed commands without any locks held, due to the handler being implicitely serialized and there being no need to look at any shared data. Only grab the lock to return the command structure to the free pool. A small optimization can still be made to collect all of the completions together and then free them together under a single lock. Items 3 and 4 significantly increase the performance of the driver. On an LSI 320-2X card, transactions per second went from 13,000 to 31,000 in my testing with these changes. However, these changes are still fairly experimental and shouldn't be merged to 6.x until there is more testing. Thanks to Doug Ambrosko, LSI, Dell, and Yahoo for contributing towards this.
2005-12-14 03:26:49 +00:00
/*
* Command for random deletion of logical drives
*/
#define FC_DEL_LOGDRV 0xA4
#define OP_SUP_DEL_LOGDRV 0x2A
#define OP_GET_LDID_MAP 0x18
#define OP_DEL_LOGDRV 0x1C
/*
* Command results
*/
#define AMR_STATUS_SUCCESS 0x00
#define AMR_STATUS_ABORTED 0x02
#define AMR_STATUS_FAILED 0x80
/*
* Physical/logical drive states
*/
#define AMR_DRV_CURSTATE(x) ((x) & 0x0f)
#define AMR_DRV_PREVSTATE(x) (((x) >> 4) & 0x0f)
#define AMR_DRV_OFFLINE 0x00
#define AMR_DRV_DEGRADED 0x01
#define AMR_DRV_OPTIMAL 0x02
#define AMR_DRV_ONLINE 0x03
#define AMR_DRV_FAILED 0x04
#define AMR_DRV_REBUILD 0x05
#define AMR_DRV_HOTSPARE 0x06
/*
* Logical drive properties
*/
#define AMR_DRV_RAID_MASK 0x0f /* RAID level 0, 1, 3, 5, etc. */
#define AMR_DRV_WRITEBACK 0x10 /* write-back enabled */
#define AMR_DRV_READHEAD 0x20 /* readhead policy enabled */
#define AMR_DRV_ADAPTIVE 0x40 /* adaptive I/O policy enabled */
/*
* Battery status
*/
#define AMR_BATT_MODULE_MISSING 0x01
#define AMR_BATT_LOW_VOLTAGE 0x02
#define AMR_BATT_TEMP_HIGH 0x04
#define AMR_BATT_PACK_MISSING 0x08
#define AMR_BATT_CHARGE_MASK 0x30
#define AMR_BATT_CHARGE_DONE 0x00
#define AMR_BATT_CHARGE_INPROG 0x10
#define AMR_BATT_CHARGE_FAIL 0x20
#define AMR_BATT_CYCLES_EXCEEDED 0x40
/********************************************************************************
********************************************************************************
8LD Firmware Interface
********************************************************************************
********************************************************************************/
/*
* Array constraints
*/
#define AMR_8LD_MAXDRIVES 8
#define AMR_8LD_MAXCHAN 5
#define AMR_8LD_MAXTARG 15
#define AMR_8LD_MAXPHYSDRIVES (AMR_8LD_MAXCHAN * AMR_8LD_MAXTARG)
/*
* Adapter Info structure
*/
struct amr_adapter_info
{
u_int8_t aa_maxio;
u_int8_t aa_rebuild_rate;
u_int8_t aa_maxtargchan;
u_int8_t aa_channels;
u_int8_t aa_firmware[4];
u_int16_t aa_flashage;
u_int8_t aa_chipsetvalue;
u_int8_t aa_memorysize;
u_int8_t aa_cacheflush;
u_int8_t aa_bios[4];
u_int8_t aa_boardtype;
u_int8_t aa_scsisensealert;
u_int8_t aa_writeconfigcount;
u_int8_t aa_driveinsertioncount;
u_int8_t aa_inserteddrive;
u_int8_t aa_batterystatus;
u_int8_t res1;
2002-09-23 18:54:32 +00:00
} __packed;
/*
* Logical Drive info structure
*/
struct amr_logdrive_info
{
u_int8_t al_numdrives;
u_int8_t res1[3];
u_int32_t al_size[AMR_8LD_MAXDRIVES];
u_int8_t al_properties[AMR_8LD_MAXDRIVES];
u_int8_t al_state[AMR_8LD_MAXDRIVES];
2002-09-23 18:54:32 +00:00
} __packed;
/*
* Physical Drive info structure
*/
struct amr_physdrive_info
{
u_int8_t ap_state[AMR_8LD_MAXPHYSDRIVES]; /* low nibble current state, high nibble previous state */
u_int8_t ap_predictivefailure;
2002-09-23 18:54:32 +00:00
} __packed;
/*
* Enquiry response structure for AMR_CMD_ENQUIRY, AMR_CMD_EXT_ENQUIRY and
* AMR_CMD_EXT_ENQUIRY2.
* ENQUIRY EXT_ENQUIRY EXT_ENQUIRY2
*/
struct amr_enquiry
{
struct amr_adapter_info ae_adapter; /* X X X */
struct amr_logdrive_info ae_ldrv; /* X X X */
struct amr_physdrive_info ae_pdrv; /* X X X */
u_int8_t ae_formatting[AMR_8LD_MAXDRIVES];/* X X */
u_int8_t res1[AMR_8LD_MAXDRIVES]; /* X X */
u_int32_t ae_extlen; /* X */
u_int16_t ae_subsystem; /* X */
u_int16_t ae_subvendor; /* X */
u_int32_t ae_signature; /* X */
#define AMR_SIG_431 0xfffe0001
#define AMR_SIG_438 0xfffd0002
#define AMR_SIG_762 0xfffc0003
#define AMR_SIG_T5 0xfffb0004
#define AMR_SIG_466 0xfffa0005
#define AMR_SIG_467 0xfff90006
#define AMR_SIG_T7 0xfff80007
#define AMR_SIG_490 0xfff70008
u_int8_t res2[844]; /* X */
2002-09-23 18:54:32 +00:00
} __packed;
/********************************************************************************
********************************************************************************
40LD Firmware Interface
********************************************************************************
********************************************************************************/
/*
* Array constraints
*/
#define AMR_40LD_MAXDRIVES 40
#define AMR_40LD_MAXCHAN 16
#define AMR_40LD_MAXTARG 16
#define AMR_40LD_MAXPHYSDRIVES 256
/*
* Product Info structure
*/
struct amr_prodinfo
{
u_int32_t ap_size; /* current size in bytes (not including resvd) */
u_int32_t ap_configsig; /* default is 0x00282008, indicating 0x28 maximum
* logical drives, 0x20 maximum stripes and 0x08
* maximum spans */
u_int8_t ap_firmware[16]; /* printable identifiers */
u_int8_t ap_bios[16];
u_int8_t ap_product[80];
u_int8_t ap_maxio; /* maximum number of concurrent commands supported */
u_int8_t ap_nschan; /* number of SCSI channels present */
u_int8_t ap_fcloops; /* number of fibre loops present */
u_int8_t ap_memtype; /* memory type */
u_int32_t ap_signature;
u_int16_t ap_memsize; /* onboard memory in MB */
u_int16_t ap_subsystem; /* subsystem identifier */
u_int16_t ap_subvendor; /* subsystem vendor ID */
u_int8_t ap_numnotifyctr; /* number of notify counters */
2002-09-23 18:54:32 +00:00
} __packed;
/*
* Notify structure
*/
struct amr_notify
{
u_int32_t an_globalcounter; /* change counter */
u_int8_t an_paramcounter; /* parameter change counter */
u_int8_t an_paramid;
#define AMR_PARAM_REBUILD_RATE 0x01 /* value = new rebuild rate */
#define AMR_PARAM_FLUSH_INTERVAL 0x02 /* value = new flush interval */
#define AMR_PARAM_SENSE_ALERT 0x03 /* value = last physical drive with check condition set */
#define AMR_PARAM_DRIVE_INSERTED 0x04 /* value = last physical drive inserted */
#define AMR_PARAM_BATTERY_STATUS 0x05 /* value = battery status */
u_int16_t an_paramval;
u_int8_t an_writeconfigcounter; /* write config occurred */
u_int8_t res1[3];
u_int8_t an_ldrvopcounter; /* logical drive operation started/completed */
u_int8_t an_ldrvopid;
u_int8_t an_ldrvopcmd;
#define AMR_LDRVOP_CHECK 0x01
#define AMR_LDRVOP_INIT 0x02
#define AMR_LDRVOP_REBUILD 0x03
u_int8_t an_ldrvopstatus;
#define AMR_LDRVOP_SUCCESS 0x00
#define AMR_LDRVOP_FAILED 0x01
#define AMR_LDRVOP_ABORTED 0x02
#define AMR_LDRVOP_CORRECTED 0x03
#define AMR_LDRVOP_STARTED 0x04
u_int8_t an_ldrvstatecounter; /* logical drive state change occurred */
u_int8_t an_ldrvstateid;
u_int8_t an_ldrvstatenew;
u_int8_t an_ldrvstateold;
u_int8_t an_pdrvstatecounter; /* physical drive state change occurred */
u_int8_t an_pdrvstateid;
u_int8_t an_pdrvstatenew;
u_int8_t an_pdrvstateold;
u_int8_t an_pdrvfmtcounter;
u_int8_t an_pdrvfmtid;
u_int8_t an_pdrvfmtval;
#define AMR_FORMAT_START 0x01
#define AMR_FORMAT_COMPLETE 0x02
u_int8_t res2;
u_int8_t an_targxfercounter; /* scsi xfer rate change */
u_int8_t an_targxferid;
u_int8_t an_targxferval;
u_int8_t res3;
u_int8_t an_fcloopidcounter; /* FC/AL loop ID changed */
u_int8_t an_fcloopidpdrvid;
u_int8_t an_fcloopid0;
u_int8_t an_fcloopid1;
u_int8_t an_fcloopstatecounter; /* FC/AL loop status changed */
u_int8_t an_fcloopstate0;
u_int8_t an_fcloopstate1;
u_int8_t res4;
2002-09-23 18:54:32 +00:00
} __packed;
/*
* Enquiry3 structure
*/
struct amr_enquiry3
{
u_int32_t ae_datasize; /* valid data size in this structure */
union { /* event notify structure */
struct amr_notify n;
u_int8_t pad[0x80];
} ae_notify;
u_int8_t ae_rebuildrate; /* current rebuild rate in % */
u_int8_t ae_cacheflush; /* flush interval in seconds */
u_int8_t ae_sensealert;
u_int8_t ae_driveinsertcount; /* count of inserted drives */
u_int8_t ae_batterystatus;
u_int8_t ae_numldrives;
u_int8_t ae_reconstate[AMR_40LD_MAXDRIVES / 8]; /* reconstruction state */
u_int16_t ae_opstatus[AMR_40LD_MAXDRIVES / 8]; /* operation status per drive */
u_int32_t ae_drivesize[AMR_40LD_MAXDRIVES]; /* logical drive size */
u_int8_t ae_driveprop[AMR_40LD_MAXDRIVES]; /* logical drive properties */
u_int8_t ae_drivestate[AMR_40LD_MAXDRIVES]; /* logical drive state */
u_int8_t ae_pdrivestate[AMR_40LD_MAXPHYSDRIVES]; /* physical drive state */
u_int16_t ae_pdriveformat[AMR_40LD_MAXPHYSDRIVES / 16];
u_int8_t ae_targxfer[80]; /* physical drive transfer rates */
u_int8_t res1[263]; /* pad to 1024 bytes */
2002-09-23 18:54:32 +00:00
} __packed;
/********************************************************************************
********************************************************************************
Mailbox and Command Structures
********************************************************************************
********************************************************************************/
#define AMR_MBOX_CMDSIZE 0x10 /* portion worth copying for controller */
struct amr_mailbox
{
u_int8_t mb_command;
u_int8_t mb_ident;
Mega update to the LSI MegaRAID driver: 1. Implement a large set of ioctl shims so that the Linux management apps from LSI will work. This includes infrastructure to support adding, deleting and rescanning arrays at runtime. This is based on work from Doug Ambrosko, heavily augmented by LSI and Yahoo. 2. Implement full 64-bit DMA support. Systems with more than 4GB of RAM can now operate without the cost of bounce buffers. Cards that cannot do 64-bit DMA will automatically revert to using bounce buffers. This option can be forced off by setting the 'hw.amr.force_sg32" tunable in the loader. It should only be turned off for debugging purposes. This work was sponsored by Yahoo. 3. Streamline the command delivery and interrupt handler paths after much discussion with Dell and LSI. The logic now closely matches the intended design, making it both more robust and much faster. Certain i/o failures under heavy load should be fixed with this. 4. Optimize the locking. In the interrupt handler, the card can be checked for completed commands without any locks held, due to the handler being implicitely serialized and there being no need to look at any shared data. Only grab the lock to return the command structure to the free pool. A small optimization can still be made to collect all of the completions together and then free them together under a single lock. Items 3 and 4 significantly increase the performance of the driver. On an LSI 320-2X card, transactions per second went from 13,000 to 31,000 in my testing with these changes. However, these changes are still fairly experimental and shouldn't be merged to 6.x until there is more testing. Thanks to Doug Ambrosko, LSI, Dell, and Yahoo for contributing towards this.
2005-12-14 03:26:49 +00:00
u_int16_t mb_blkcount; /* u_int8_t opcode */
/* u_int8_t subopcode */
u_int32_t mb_lba;
u_int32_t mb_physaddr;
u_int8_t mb_drive;
Mega update to the LSI MegaRAID driver: 1. Implement a large set of ioctl shims so that the Linux management apps from LSI will work. This includes infrastructure to support adding, deleting and rescanning arrays at runtime. This is based on work from Doug Ambrosko, heavily augmented by LSI and Yahoo. 2. Implement full 64-bit DMA support. Systems with more than 4GB of RAM can now operate without the cost of bounce buffers. Cards that cannot do 64-bit DMA will automatically revert to using bounce buffers. This option can be forced off by setting the 'hw.amr.force_sg32" tunable in the loader. It should only be turned off for debugging purposes. This work was sponsored by Yahoo. 3. Streamline the command delivery and interrupt handler paths after much discussion with Dell and LSI. The logic now closely matches the intended design, making it both more robust and much faster. Certain i/o failures under heavy load should be fixed with this. 4. Optimize the locking. In the interrupt handler, the card can be checked for completed commands without any locks held, due to the handler being implicitely serialized and there being no need to look at any shared data. Only grab the lock to return the command structure to the free pool. A small optimization can still be made to collect all of the completions together and then free them together under a single lock. Items 3 and 4 significantly increase the performance of the driver. On an LSI 320-2X card, transactions per second went from 13,000 to 31,000 in my testing with these changes. However, these changes are still fairly experimental and shouldn't be merged to 6.x until there is more testing. Thanks to Doug Ambrosko, LSI, Dell, and Yahoo for contributing towards this.
2005-12-14 03:26:49 +00:00
u_int8_t mb_nsgelem; /* u_int8_t rserv[0] */
u_int8_t res1; /* u_int8_t rserv[1] */
u_int8_t mb_busy; /* u_int8_t rserv[2] */
u_int8_t mb_nstatus;
u_int8_t mb_status;
u_int8_t mb_completed[46];
u_int8_t mb_poll;
u_int8_t mb_ack;
u_int8_t res2[16];
2002-09-23 18:54:32 +00:00
} __packed;
struct amr_mailbox64
{
Mega update to the LSI MegaRAID driver: 1. Implement a large set of ioctl shims so that the Linux management apps from LSI will work. This includes infrastructure to support adding, deleting and rescanning arrays at runtime. This is based on work from Doug Ambrosko, heavily augmented by LSI and Yahoo. 2. Implement full 64-bit DMA support. Systems with more than 4GB of RAM can now operate without the cost of bounce buffers. Cards that cannot do 64-bit DMA will automatically revert to using bounce buffers. This option can be forced off by setting the 'hw.amr.force_sg32" tunable in the loader. It should only be turned off for debugging purposes. This work was sponsored by Yahoo. 3. Streamline the command delivery and interrupt handler paths after much discussion with Dell and LSI. The logic now closely matches the intended design, making it both more robust and much faster. Certain i/o failures under heavy load should be fixed with this. 4. Optimize the locking. In the interrupt handler, the card can be checked for completed commands without any locks held, due to the handler being implicitely serialized and there being no need to look at any shared data. Only grab the lock to return the command structure to the free pool. A small optimization can still be made to collect all of the completions together and then free them together under a single lock. Items 3 and 4 significantly increase the performance of the driver. On an LSI 320-2X card, transactions per second went from 13,000 to 31,000 in my testing with these changes. However, these changes are still fairly experimental and shouldn't be merged to 6.x until there is more testing. Thanks to Doug Ambrosko, LSI, Dell, and Yahoo for contributing towards this.
2005-12-14 03:26:49 +00:00
u_int8_t pad[8]; /* Needed for alignment */
u_int32_t sg64_lo; /* S/G pointer for 64-bit commands */
u_int32_t sg64_hi; /* S/G pointer for 64-bit commands */
struct amr_mailbox mb;
2002-09-23 18:54:32 +00:00
} __packed;
struct amr_mailbox_ioctl
{
u_int8_t mb_command;
u_int8_t mb_ident;
u_int8_t mb_channel;
u_int8_t mb_param;
u_int8_t mb_pad[4];
u_int32_t mb_physaddr;
u_int8_t mb_drive;
u_int8_t mb_nsgelem;
u_int8_t res1;
u_int8_t mb_busy;
u_int8_t mb_nstatus;
u_int8_t mb_completed[46];
u_int8_t mb_poll;
u_int8_t mb_ack;
u_int8_t res4[16];
2002-09-23 18:54:32 +00:00
} __packed;
struct amr_sgentry
{
u_int32_t sg_addr;
u_int32_t sg_count;
2002-09-23 18:54:32 +00:00
} __packed;
Mega update to the LSI MegaRAID driver: 1. Implement a large set of ioctl shims so that the Linux management apps from LSI will work. This includes infrastructure to support adding, deleting and rescanning arrays at runtime. This is based on work from Doug Ambrosko, heavily augmented by LSI and Yahoo. 2. Implement full 64-bit DMA support. Systems with more than 4GB of RAM can now operate without the cost of bounce buffers. Cards that cannot do 64-bit DMA will automatically revert to using bounce buffers. This option can be forced off by setting the 'hw.amr.force_sg32" tunable in the loader. It should only be turned off for debugging purposes. This work was sponsored by Yahoo. 3. Streamline the command delivery and interrupt handler paths after much discussion with Dell and LSI. The logic now closely matches the intended design, making it both more robust and much faster. Certain i/o failures under heavy load should be fixed with this. 4. Optimize the locking. In the interrupt handler, the card can be checked for completed commands without any locks held, due to the handler being implicitely serialized and there being no need to look at any shared data. Only grab the lock to return the command structure to the free pool. A small optimization can still be made to collect all of the completions together and then free them together under a single lock. Items 3 and 4 significantly increase the performance of the driver. On an LSI 320-2X card, transactions per second went from 13,000 to 31,000 in my testing with these changes. However, these changes are still fairly experimental and shouldn't be merged to 6.x until there is more testing. Thanks to Doug Ambrosko, LSI, Dell, and Yahoo for contributing towards this.
2005-12-14 03:26:49 +00:00
struct amr_sg64entry
{
u_int64_t sg_addr;
u_int32_t sg_count;
} __packed;
struct amr_passthrough
{
u_int8_t ap_timeout:3;
u_int8_t ap_ars:1;
u_int8_t ap_dummy:3;
u_int8_t ap_islogical:1;
u_int8_t ap_logical_drive_no;
u_int8_t ap_channel;
u_int8_t ap_scsi_id;
u_int8_t ap_queue_tag;
u_int8_t ap_queue_action;
u_int8_t ap_cdb[AMR_MAX_CDB_LEN];
u_int8_t ap_cdb_length;
u_int8_t ap_request_sense_length;
u_int8_t ap_request_sense_area[AMR_MAX_REQ_SENSE_LEN];
u_int8_t ap_no_sg_elements;
u_int8_t ap_scsi_status;
u_int32_t ap_data_transfer_address;
u_int32_t ap_data_transfer_length;
2002-09-23 18:54:32 +00:00
} __packed;
struct amr_ext_passthrough
{
u_int8_t ap_timeout:3;
u_int8_t ap_ars:1;
u_int8_t ap_rsvd1:1;
u_int8_t ap_cd_rom:1;
u_int8_t ap_rsvd2:1;
u_int8_t ap_islogical:1;
u_int8_t ap_logical_drive_no;
u_int8_t ap_channel;
u_int8_t ap_scsi_id;
u_int8_t ap_queue_tag;
u_int8_t ap_queue_action;
u_int8_t ap_cdb_length;
u_int8_t ap_rsvd3;
u_int8_t ap_cdb[AMR_MAX_EXTCDB_LEN];
u_int8_t ap_no_sg_elements;
u_int8_t ap_scsi_status;
u_int8_t ap_request_sense_length;
u_int8_t ap_request_sense_area[AMR_MAX_REQ_SENSE_LEN];
u_int8_t ap_rsvd4;
u_int32_t ap_data_transfer_address;
u_int32_t ap_data_transfer_length;
} __packed;
Mega update to the LSI MegaRAID driver: 1. Implement a large set of ioctl shims so that the Linux management apps from LSI will work. This includes infrastructure to support adding, deleting and rescanning arrays at runtime. This is based on work from Doug Ambrosko, heavily augmented by LSI and Yahoo. 2. Implement full 64-bit DMA support. Systems with more than 4GB of RAM can now operate without the cost of bounce buffers. Cards that cannot do 64-bit DMA will automatically revert to using bounce buffers. This option can be forced off by setting the 'hw.amr.force_sg32" tunable in the loader. It should only be turned off for debugging purposes. This work was sponsored by Yahoo. 3. Streamline the command delivery and interrupt handler paths after much discussion with Dell and LSI. The logic now closely matches the intended design, making it both more robust and much faster. Certain i/o failures under heavy load should be fixed with this. 4. Optimize the locking. In the interrupt handler, the card can be checked for completed commands without any locks held, due to the handler being implicitely serialized and there being no need to look at any shared data. Only grab the lock to return the command structure to the free pool. A small optimization can still be made to collect all of the completions together and then free them together under a single lock. Items 3 and 4 significantly increase the performance of the driver. On an LSI 320-2X card, transactions per second went from 13,000 to 31,000 in my testing with these changes. However, these changes are still fairly experimental and shouldn't be merged to 6.x until there is more testing. Thanks to Doug Ambrosko, LSI, Dell, and Yahoo for contributing towards this.
2005-12-14 03:26:49 +00:00
struct amr_linux_ioctl {
u_int32_t inlen;
u_int32_t outlen;
union {
u_int8_t fca[16];
struct {
u_int8_t opcode;
u_int8_t subopcode;
u_int16_t adapno;
u_int32_t buffer;
u_int8_t pad[4];
u_int32_t length;
} __packed fcs;
} __packed ui;
u_int8_t mbox[18];
struct amr_passthrough pthru;
u_int32_t data;
u_int8_t pad[4];
} __packed;
#ifdef _KERNEL
/********************************************************************************
********************************************************************************
"Quartz" i960 PCI bridge interface
********************************************************************************
********************************************************************************/
#define AMR_CFG_SIG 0xa0 /* PCI config register for signature */
#define AMR_SIGNATURE_1 0xCCCC /* i960 signature (older adapters) */
#define AMR_SIGNATURE_2 0x3344 /* i960 signature (newer adapters) */
/*
* Doorbell registers
*/
#define AMR_QIDB 0x20
#define AMR_QODB 0x2c
#define AMR_QIDB_SUBMIT 0x00000001 /* mailbox ready for work */
#define AMR_QIDB_ACK 0x00000002 /* mailbox done */
#define AMR_QODB_READY 0x10001234 /* work ready to be processed */
/*
* Initialisation status
*/
#define AMR_QINIT_SCAN 0x01 /* init scanning drives */
#define AMR_QINIT_SCANINIT 0x02 /* init scanning initialising */
#define AMR_QINIT_FIRMWARE 0x03 /* init firmware initing */
#define AMR_QINIT_INPROG 0xdc /* init in progress */
#define AMR_QINIT_SPINUP 0x2c /* init spinning drives */
#define AMR_QINIT_NOMEM 0xac /* insufficient memory */
#define AMR_QINIT_CACHEFLUSH 0xbc /* init flushing cache */
#define AMR_QINIT_DONE 0x9c /* init successfully done */
/*
* I/O primitives
*/
#define AMR_QPUT_IDB(sc, val) bus_space_write_4(sc->amr_btag, sc->amr_bhandle, AMR_QIDB, val)
#define AMR_QGET_IDB(sc) bus_space_read_4 (sc->amr_btag, sc->amr_bhandle, AMR_QIDB)
#define AMR_QPUT_ODB(sc, val) bus_space_write_4(sc->amr_btag, sc->amr_bhandle, AMR_QODB, val)
#define AMR_QGET_ODB(sc) bus_space_read_4 (sc->amr_btag, sc->amr_bhandle, AMR_QODB)
#ifdef AMR_BOARD_INIT
#define AMR_QRESET(sc) \
do { \
pci_write_config((sc)->amr_dev, 0x40, pci_read_config((sc)->amr_dev, 0x40, 1) | 0x20, 1); \
pci_write_config((sc)->amr_dev, 0x64, 0x1122, 1); \
} while (0)
#define AMR_QGET_INITSTATUS(sc) pci_read_config((sc)->amr_dev, 0x9c, 1)
#define AMR_QGET_INITCHAN(sc) pci_read_config((sc)->amr_dev, 0x9f, 1)
#define AMR_QGET_INITTARG(sc) pci_read_config((sc)->amr_dev, 0x9e, 1)
#endif
/********************************************************************************
********************************************************************************
"Standard" old-style ASIC bridge interface
********************************************************************************
********************************************************************************/
/*
* I/O registers
*/
#define AMR_SCMD 0x10 /* command/ack register (write) */
#define AMR_SMBOX_BUSY 0x10 /* mailbox status (read) */
#define AMR_STOGGLE 0x11 /* interrupt enable bit here */
#define AMR_SMBOX_0 0x14 /* mailbox physical address low byte */
#define AMR_SMBOX_1 0x15
#define AMR_SMBOX_2 0x16
#define AMR_SMBOX_3 0x17 /* high byte */
#define AMR_SMBOX_ENABLE 0x18 /* atomic mailbox address enable */
#define AMR_SINTR 0x1a /* interrupt status */
/*
* I/O magic numbers
*/
#define AMR_SCMD_POST 0x10 /* -> SCMD to initiate action on mailbox */
#define AMR_SCMD_ACKINTR 0x08 /* -> SCMD to ack mailbox retrieved */
#define AMR_STOGL_IENABLE 0xc0 /* in STOGGLE */
#define AMR_SINTR_VALID 0x40 /* in SINTR */
#define AMR_SMBOX_BUSYFLAG 0x10 /* in SMBOX_BUSY */
#define AMR_SMBOX_ADDR 0x00 /* -> SMBOX_ENABLE */
/*
* Initialisation status
*/
#define AMR_SINIT_ABEND 0xee /* init abnormal terminated */
#define AMR_SINIT_NOMEM 0xca /* insufficient memory */
#define AMR_SINIT_CACHEFLUSH 0xbb /* firmware flushing cache */
#define AMR_SINIT_INPROG 0x11 /* init in progress */
#define AMR_SINIT_SPINUP 0x22 /* firmware spinning drives */
#define AMR_SINIT_DONE 0x99 /* init successfully done */
/*
* I/O primitives
*/
#define AMR_SPUT_ISTAT(sc, val) bus_space_write_1(sc->amr_btag, sc->amr_bhandle, AMR_SINTR, val)
#define AMR_SGET_ISTAT(sc) bus_space_read_1 (sc->amr_btag, sc->amr_bhandle, AMR_SINTR)
#define AMR_SACK_INTERRUPT(sc) bus_space_write_1(sc->amr_btag, sc->amr_bhandle, AMR_SCMD, AMR_SCMD_ACKINTR)
#define AMR_SPOST_COMMAND(sc) bus_space_write_1(sc->amr_btag, sc->amr_bhandle, AMR_SCMD, AMR_SCMD_POST)
#define AMR_SGET_MBSTAT(sc) bus_space_read_1 (sc->amr_btag, sc->amr_bhandle, AMR_SMBOX_BUSY)
#define AMR_SENABLE_INTR(sc) \
bus_space_write_1(sc->amr_btag, sc->amr_bhandle, AMR_STOGGLE, \
bus_space_read_1(sc->amr_btag, sc->amr_bhandle, AMR_STOGGLE) | AMR_STOGL_IENABLE)
#define AMR_SDISABLE_INTR(sc) \
bus_space_write_1(sc->amr_btag, sc->amr_bhandle, AMR_STOGGLE, \
bus_space_read_1(sc->amr_btag, sc->amr_bhandle, AMR_STOGGLE) & ~AMR_STOGL_IENABLE)
#define AMR_SBYTE_SET(sc, reg, val) bus_space_write_1(sc->amr_btag, sc->amr_bhandle, reg, val)
#ifdef AMR_BOARD_INIT
#define AMR_SRESET(sc) bus_space_write_1(sc->amr_btag, sc->amr_bhandle, 0, 0x80)
#define AMR_SGET_INITSTATUS(sc) bus_space_read_1 (sc->amr_btag, sc->amr_bhandle, AMR_SMBOX_ENABLE)
#define AMR_SGET_FAILDRIVE(sc) bus_space_read_1 (sc->amr_btag, sc->amr_bhandle, AMR_SMBOX_ENABLE + 1)
#define AMR_SGET_INITCHAN(sc) bus_space_read_1 (sc->amr_btag, sc->amr_bhandle, AMR_SMBOX_ENABLE + 2)
#define AMR_SGET_INITTARG(sc) bus_space_read_1 (sc->amr_btag, sc->amr_bhandle, AMR_SMBOX_ENABLE + 3)
#endif
#endif /* _KERNEL */