This patch contains SCST core's README. Signed-off-by: Vladislav Bolkhovitin <vst@xxxxxxxx> --- README.scst | 1379 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 1379 insertions(+) diff -uprN orig/linux-2.6.33/Documentation/scst/README.scst linux-2.6.33/Documentation/scst/README.scst --- orig/linux-2.6.33/Documentation/scst/README.scst +++ linux-2.6.33/Documentation/scst/README.scst @@ -0,0 +1,1379 @@ +Generic SCSI target mid-level for Linux (SCST) +============================================== + +SCST is designed to provide unified, consistent interface between SCSI +target drivers and Linux kernel and simplify target drivers development +as much as possible. Detail description of SCST's features and internals +could be found in "Generic SCSI Target Middle Level for Linux" document +SCST's Internet page http://scst.sourceforge.net. + +SCST supports the following I/O modes: + + * Pass-through mode with one to many relationship, i.e. when multiple + initiators can connect to the exported pass-through devices, for + the following SCSI devices types: disks (type 0), tapes (type 1), + processors (type 3), CDROMs (type 5), MO disks (type 7), medium + changers (type 8) and RAID controllers (type 0xC) + + * FILEIO mode, which allows to use files on file systems or block + devices as virtual remotely available SCSI disks or CDROMs with + benefits of the Linux page cache + + * BLOCKIO mode, which performs direct block IO with a block device, + bypassing page-cache for all operations. This mode works ideally with + high-end storage HBAs and for applications that either do not need + caching between application and disk or need the large block + throughput + + * User space mode using scst_user device handler, which allows to + implement in the user space virtual SCSI devices in the SCST + environment + + * "Performance" device handlers, which provide in pseudo pass-through + mode a way for direct performance measurements without overhead of + actual data transferring from/to underlying SCSI device + +In addition, SCST supports advanced per-initiator access and devices +visibility management, so different initiators could see different set +of devices with different access permissions. See below for details. + +Installation +------------ + +To see your devices remotely, you need to add them to at least "Default" +security group (see below how). By default, no local devices are seen +remotely. There must be LUN 0 in each security group, i.e. LUs +numeration must not start from, e.g., 1. Otherwise you will see no +devices on remote initiators and SCST core will write into the kernel +log message: "tgt_dev for LUN 0 not found, command to unexisting LU?" + +It is highly recommended to use scstadmin utility for configuring +devices and security groups. + +If you experience problems during modules load or running, check your +kernel logs (or run dmesg command for the few most recent messages). + +IMPORTANT: Without loading appropriate device handler, corresponding devices +========= will be invisible for remote initiators, which could lead to holes + in the LUN addressing, so automatic device scanning by remote SCSI + mid-level could not notice the devices. Therefore you will have + to add them manually via + 'echo "- - -" >/sys/class/scsi_host/hostX/scan', + where X - is the host number. + +IMPORTANT: Working of target and initiator on the same host is +========= supported, except the following 2 cases: swap over target exported + device and using a writable mmap over a file from target + exported device. The latter means you can't mount a file + system over target exported device. In other words, you can + freely use any sg, sd, st, etc. devices imported from target + on the same host, but you can't mount file systems or put + swap on them. This is a limitation of Linux memory/cache + manager, because in this case an OOM deadlock like: system + needs some memory -> it decides to clear some cache -> cache + needs to write on target exported device -> initiator sends + request to the target -> target needs memory -> system needs + even more memory -> deadlock. + +IMPORTANT: In the current version simultaneous access to local SCSI devices +========= via standard high-level SCSI drivers (sd, st, sg, etc.) and + SCST's target drivers is unsupported. Especially it is + important for execution via sg and st commands that change + the state of devices and their parameters, because that could + lead to data corruption. If any such command is done, at + least related device handler(s) must be restarted. For block + devices READ/WRITE commands using direct disk handler look to + be safe. + +IMPORTANT: Some versions of Windows have a bug, which makes them consider +========= response of READ CAPACITY(16) longer than 12 bytes as a faulty one. + As the result, such Windows'es refuse to see SCST exported + devices >2TB in size. This is fixed by MS in latter Windows + versions, probably, by some hotfix. But if you're using such + buggy Windows and experience this problem, change in + scst_vdisk.c::vdisk_exec_read_capacity16() "#if 1" to "#if 0". + +Usage in failover mode +---------------------- + +It is recommended to use TEST UNIT READY ("tur") command to check if +SCST target is alive in MPIO configurations. + +Device handlers +--------------- + +Device specific drivers (device handlers) are plugins for SCST, which +help SCST to analyze incoming requests and determine parameters, +specific to various types of devices. If an appropriate device handler +for a SCSI device type isn't loaded, SCST doesn't know how to handle +devices of this type, so they will be invisible for remote initiators +(more precisely, "LUN not supported" sense code will be returned). + +In addition to device handlers for real devices, there are VDISK, user +space and "performance" device handlers. + +VDISK device handler works over files on file systems and makes from +them virtual remotely available SCSI disks or CDROM's. In addition, it +allows to work directly over a block device, e.g. local IDE or SCSI disk +or ever disk partition, where there is no file systems overhead. Using +block devices comparing to sending SCSI commands directly to SCSI +mid-level via scsi_do_req()/scsi_execute_async() has advantage that data +are transferred via system cache, so it is possible to fully benefit from +caching and read ahead performed by Linux's VM subsystem. The only +disadvantage here that in the FILEIO mode there is superfluous data +copying between the cache and SCST's buffers. This issue is going to be +addressed in the next release. Virtual CDROM's are useful for remote +installation. See below for details how to setup and use VDISK device +handler. + +SCST user space device handler provides an interface between SCST and +the user space, which allows to create pure user space devices. The +simplest example, where one would want it is if he/she wants to write a +VTL. With scst_user he/she can write it purely in the user space. Or one +would want it if he/she needs some sophisticated for kernel space +processing of the passed data, like encrypting them or making snapshots. + +"Performance" device handlers for disks, MO disks and tapes in their +exec() method skip (pretend to execute) all READ and WRITE operations +and thus provide a way for direct link performance measurements without +overhead of actual data transferring from/to underlying SCSI device. + +NOTE: Since "perf" device handlers on READ operations don't touch the +==== commands' data buffer, it is returned to remote initiators as it + was allocated, without even being zeroed. Thus, "perf" device + handlers impose some security risk, so use them with caution. + +Compilation options +------------------- + +There are the following compilation options, that could be change using +your favorite kernel configuration Makefile target, e.g. "make xconfig": + + - CONFIG_SCST_DEBUG - if defined, turns on some debugging code, + including some logging. Makes the driver considerably bigger and slower, + producing large amount of log data. + + - CONFIG_SCST_TRACING - if defined, turns on ability to log events. Makes the + driver considerably bigger and leads to some performance loss. + + - CONFIG_SCST_EXTRACHECKS - if defined, adds extra validity checks in + the various places. + + - CONFIG_SCST_USE_EXPECTED_VALUES - if not defined (default), initiator + supplied expected data transfer length and direction will be used only for + verification purposes to return error or warn in case if one of them + is invalid. Instead, locally decoded from SCSI command values will be + used. This is necessary for security reasons, because otherwise a + faulty initiator can crash target by supplying invalid value in one + of those parameters. This is especially important in case of + pass-through mode. If CONFIG_SCST_USE_EXPECTED_VALUES is defined, initiator + supplied expected data transfer length and direction will override + the locally decoded values. This might be necessary if internal SCST + commands translation table doesn't contain SCSI command, which is + used in your environment. You can know that if you have messages like + "Unknown opcode XX for YY. Should you update scst_scsi_op_table?" in + your kernel log and your initiator returns an error. Also report + those messages in the SCST mailing list + scst-devel@xxxxxxxxxxxxxxxxxxxxxx Note, that not all SCSI transports + support supplying expected values. + + - CONFIG_SCST_DEBUG_TM - if defined, turns on task management functions + debugging, when on LUN 6 some of the commands will be delayed for + about 60 sec., so making the remote initiator send TM functions, eg + ABORT TASK and TARGET RESET. Also define + CONFIG_SCST_TM_DBG_GO_OFFLINE symbol in the Makefile if you want that + the device eventually become completely unresponsive, or otherwise to + circle around ABORTs and RESETs code. Needs CONFIG_SCST_DEBUG turned + on. + + - CONFIG_SCST_STRICT_SERIALIZING - if defined, makes SCST send all commands to + underlying SCSI device synchronously, one after one. This makes task + management more reliable, with cost of some performance penalty. This + is mostly actual for stateful SCSI devices like tapes, where the + result of command's execution depends from device's settings defined + by previous commands. Disk and RAID devices are stateless in the most + cases. The current SCSI core in Linux doesn't allow to abort all + commands reliably if they sent asynchronously to a stateful device. + Turned off by default, turn it on if you use stateful device(s) and + need as much error recovery reliability as possible. As a side effect + of CONFIG_SCST_STRICT_SERIALIZING, on kernels below 2.6.30 no kernel + patching is necessary for pass-through device handlers (scst_disk, + etc.). + + - CONFIG_SCST_ALLOW_PASSTHROUGH_IO_SUBMIT_IN_SIRQ - if defined, it will be + allowed to submit pass-through commands to real SCSI devices via the SCSI + middle layer using scsi_execute_async() function from soft IRQ + context (tasklets). This used to be the default, but currently it + seems the SCSI middle layer starts expecting only thread context on + the IO submit path, so it is disabled now by default. Enabling it + will decrease amount of context switches and improve performance. It + is more or less safe, in the worst case, if in your configuration the + SCSI middle layer really doesn't expect SIRQ context in + scsi_execute_async() function, you will get a warning message in the + kernel log. + + - CONFIG_SCST_STRICT_SECURITY - if defined, makes SCST zero allocated data + buffers. Undefining it (default) considerably improves performance + and eases CPU load, but could create a security hole (information + leakage), so enable it, if you have strict security requirements. + + - CONFIG_SCST_ABORT_CONSIDER_FINISHED_TASKS_AS_NOT_EXISTING - if defined, + in case when TASK MANAGEMENT function ABORT TASK is trying to abort a + command, which has already finished, remote initiator, which sent the + ABORT TASK request, will receive TASK NOT EXIST (or ABORT FAILED) + response for the ABORT TASK request. This is more logical response, + since, because the command finished, attempt to abort it failed, but + some initiators, particularly VMware iSCSI initiator, consider TASK + NOT EXIST response as if the target got crazy and try to RESET it. + Then sometimes get crazy itself. So, this option is disabled by + default. + + - CONFIG_SCST_MEASURE_LATENCY - if defined, provides in /sys/kernel/scst_tgt + and below statisctics about average commands processing latency. You + can clear already measured results by writing 0 in the corresponding + file. Note, you need a non-preemptible kernel to have correct + results. + +HIGHMEM kernel configurations are fully supported, but not recommended +for performance reasons, except for scst_user, where they are not +supported, because this module deals with user supplied memory on a +zero-copy manner. If you need to use it, consider change VMSPLIT option +or use 64-bit system configuration instead. + +For changing VMSPLIT option (CONFIG_VMSPLIT to be precise) you should in +"make menuconfig" command set the following variables: + + - General setup->Configure standard kernel features (for small systems): ON + + - General setup->Prompt for development and/or incomplete code/drivers: ON + + - Processor type and features->High Memory Support: OFF + + - Processor type and features->Memory split: according to amount of + memory you have. If it is less than 800MB, you may not touch this + option at all. + +Module parameters +----------------- + +Module scst supports the following parameters: + + - scst_threads - allows to set count of SCST's threads. By default it + is CPU count. + + - scst_max_cmd_mem - sets maximum amount of memory in Mb allowed to be + consumed by the SCST commands for data buffers at any given time. By + default it is approximately TotalMem/4. + +SCST sysfs interface +-------------------- + +Root of SCST sysfs interface is /sys/kernel/scst_tgt. It has the +following entries: + + - devices - this is a root subdirectory for all SCST devices + + - handlers - this is a root subdirectory for all SCST dev handlers + + - sgv - this is a root subdirectory for all SCST SGV caches + + - targets - this is a root subdirectory for all SCST targets + + - setup_id - allows to read and write SCST setup ID. This ID can be + used in cases, when the same SCST configuration should be installed + on several targets, but exported from those targets devices should + have different IDs and SNs. For instance, VDISK dev handler uses this + ID to generate T10 vendor specific identifier and SN of the devices. + + - threads - allows to read and set number of global SCST I/O threads. + Those threads used with async. dev handlers, for instance, vdisk + BLOCKIO or NULLIO. + + - trace_level - allows to enable and disable various tracing + facilities. See content of this file for help how to use it. + + - version - read-only attribute, which allows to see version of + SCST and enabled optional features. + +Each SCST sysfs file (attribute) can contain in the last line mark +"[key]". It is automatically added mark used to allow scstadmin to see +which attributes it should save in the config file. You can ignore it. + +"Devices" subdirectory contains subdirectories for each SCST devices. + +Content of each device's subdirectory is dev handler specific. See +documentation for your dev handlers for more info about it as well as +SysfsRules file for more info about common to all dev handlers rules. +Standard SCST dev handlers have at least the following common entries: + + - exported - subdirectory containing links to all LUNs where this + device was exported. + + - handler - if dev handler determined for this device, this link points + to it. The handler can be not set for pass-through devices. + + - threads_num - shows and allows to set number of threads in this device's + threads pool. If 0 - no threads will be created, and global SCST + threads pool will be used. If <0 - creation of the threads pool is + prohibited. + + - threads_pool_type - shows and allows to sets threads pool type. + Possible values: "per_initiator" and "shared". When the value is + "per_initiator" (default), each session from each initiator will use + separate dedicated pool of threads. When the value is "shared", all + sessions from all initiators will share the same per-device pool of + threads. Valid only if threads_num attribute >0. + + - type - SCSI type of this device + +See below for more information about other entries of this subdirectory +of the standard SCST dev handlers. + +"Handlers" subdirectory contains subdirectories for each SCST dev +handler. + +Content of each handler's subdirectory is dev handler specific. See +documentation for your dev handlers for more info about it as well as +SysfsRules file for more info about common to all dev handlers rules. +Standard SCST dev handlers have at least the following common entries: + + - mgmt - this entry allows to create virtual devices and their + attributes (for virtual devices dev handlers) or assign/unassign real + SCSI devices to/from this dev handler (for pass-through dev + handlers). + + - pass_through - if exists, it contains 1 and this dev handler is a + pass-through dev handler. + + - trace_level - allows to enable and disable various tracing + facilities. See content of this file for help how to use it. + + - type - SCSI type of devices served by this dev handler. + +See below for more information about other entries of this subdirectory +of the standard SCST dev handlers. + +"Sgv" subdirectory contains statistic information of SCST SGV caches. It +has the following entries: + + - None, one or more subdirectories for each existing SGV cache. + + - global_stats - file containing global SGV caches statistics. + +Each SGV cache's subdirectory has the following item: + + - stats - file containing statistics for this SGV caches. + +"Targets" subdirectory contains subdirectories for each SCST target. + +Content of each target's subdirectory is target specific. See +documentation for your target for more info about it as well as +SysfsRules file for more info about common to all targets rules. +Every target should have at least the following entries: + + - ini_groups - subdirectory, which contains and allows to define + initiator-oriented access control information, see below. + + - luns - subdirectory, which contains list of available LUNs in the + target-oriented access control and allows to define it, see below. + + - sessions - subdirectory containing connected to this target sessions. + + - enabled - using this attribute you can enable or disable this target/ + It allows to finish configuring it before it starts accepting new + connections. 0 by default. + + - addr_method - used LUNs addressing method. Possible values: + "Peripheral" and "Flat". Most initiators work well with Peripheral + addressing method (default), but some (HP-UX, for instance) may + require Flat method. This attribute is also available in the + initiators security groups, so you can assign the addressing method + on per-initiator basis. + + - io_grouping_type - defines how I/O from sessions to this target are + grouped together. This I/O grouping is very important for + performance. By setting this attribute in a right value, you can + considerably increase performance of your setup. This grouping is + performed only if you use CFQ I/O scheduler on the target and for + devices with threads_num >= 0 and, if threads_num > 0, with + threads_pool_type "per_initiator". Possible values: + "this_group_only", "never", "auto", or I/O group number >0. When the + value is "this_group_only" all I/O from all sessions in this target + will be grouped together. When the value is "never", I/O from + different sessions will not be grouped together, i.e. all sessions in + this target will have separate dedicated I/O groups. When the value + is "auto" (default), all I/O from initiators with the same name + (iSCSI initiator name, for instance) in all targets will be grouped + together with a separate dedicated I/O group for each initiator name. + For iSCSI this mode works well, but other transports usually use + different initiator names for different sessions, so using such + transports in MPIO configurations you should either use value + "this_group_only", or an explicit I/O group number. This attribute is + also available in the initiators security groups, so you can assign + the I/O grouping on per-initiator basis. See below for more info how + to use this attribute. + + - rel_tgt_id - allows to read or write SCSI Relative Target Port + Identifier attribute. This identifier is used to identify SCSI Target + Ports by some SCSI commands, mainly by Persistent Reservations + commands. This identifier must be unique among all SCST targets, but + for convenience SCST allows disabled targets to have not unique + rel_tgt_id. In this case SCST will not allow to enable this target + until rel_tgt_id becomes unique. This attribute initialized unique by + SCST by default. + +A target driver may have also the following entries: + + - "hw_target" - if the target driver supports both hardware and virtual + targets (for instance, an FC adapter supporting NPIV, which has + hardware targets for its physical ports as well as virtual NPIV + targets), this read only attribute for all hardware targets will + exist and contain value 1. + +Subdirectory "sessions" contains one subdirectory for each connected +session with name equal to name of the connected initiator. + +Each session subdirectory contains the following entries: + + - initiator_name - contains initiator name + + - force_close - optional write-only attribute, which allows to force + close this session. + + - active_commands - contains number of active, i.e. not yet or being + executed, SCSI commands in this session. + + - commands - contains overall number of SCSI commands in this session. + + - other target driver specific attributes and subdirectories. + +See below description of the VDISK's sysfs interface for samples. + +Access and devices visibility management (LUN masking) - sysfs interface +------------------------------------------------------------------------ + +Access and devices visibility management allows for an initiator or +group of initiators to see different devices with different LUNs +with necessary access permissions. + +SCST supports two modes of access control: + +1. Target-oriented. In this mode you define for each target a default +set of LUNs, which are accessible to all initiators, connected to that +target. This is a regular access control mode, which people usually mean +thinking about access control in general. For instance, in IET this is +the only supported mode. + +2. Initiator-oriented. In this mode you define which LUNs are accessible +for each initiator. In this mode you should create for each set of one +or more initiators, which should access to the same set of devices with +the same LUNs, a separate security group, then add to it devices and +names of allowed initiator(s). + +Both modes can be used simultaneously. In this case the +initiator-oriented mode has higher priority, than the target-oriented, +i.e. initiators are at first searched in all defined security groups for +this target and, if none matches, the default target's set of LUNs is +used. This set of LUNs might be empty, then the initiator will not see +any LUNs from the target. + +You can at any time find out which set of LUNs each session is assigned +to by looking where link +/sys/kernel/scst_tgt/targets/target_driver/target_name/sessions/initiator_name/luns +points to. + +To configure the target-oriented access control SCST provides the +following interface. Each target's sysfs subdirectory +(/sys/kernel/scst_tgt/targets/target_driver/target_name) has "luns" +subdirectory. This subdirectory contains the list of already defined +target-oriented access control LUNs for this target as well as file +"mgmt". This file has the following commands, which you can send to it, +for instance, using "echo" shell command. You can always get a small +help about supported commands by looking inside this file. "Parameters" +are one or more param_name=value pairs separated by ';'. + + - "add H:C:I:L lun [parameters]" - adds a pass-through device with + host:channel:id:lun with LUN "lun". Optionally, the device could be + marked as read only by using parameter "read_only". The recommended + way to find out H:C:I:L numbers is use of lsscsi utility. + + - "replace H:C:I:L lun [parameters]" - replaces by pass-through device + with host:channel:id:lun existing with LUN "lun" device with + generation of INQUIRY DATA HAS CHANGED Unit Attention. If the old + device doesn't exist, this command acts as the "add" command. + Optionally, the device could be marked as read only by using + parameter "read_only". The recommended way to find out H:C:I:L + numbers is use of lsscsi utility. + + - "del H:C:I:L" - deletes a pass-through device with host:channel:id:lun + The recommended way to find out H:C:I:L numbers is use of lsscsi + utility. + + - "add VNAME lun [parameters]" - adds a virtual device with name VNAME + with LUN "lun". Optionally, the device could be marked as read only + by using parameter "read_only". + + - "replace VNAME lun [parameters]" - replaces by virtual device + with name VNAME existing with LUN "lun" device with generation of + INQUIRY DATA HAS CHANGED Unit Attention. If the old device doesn't + exist, this command acts as the "add" command. Optionally, the device + could be marked as read only by using parameter "read_only". + + - "del VNAME" - deletes a virtual device with name VNAME. + + - "clear" - clears the list of devices + +To configure the initiator-oriented access control SCST provides the +following interface. Each target's sysfs subdirectory +(/sys/kernel/scst_tgt/targets/target_driver/target_name) has "ini_groups" +subdirectory. This subdirectory contains the list of already defined +security groups for this target as well as file "mgmt". This file has +the following commands, which you can send to it, for instance, using +"echo" shell command. You can always get a small help about supported +commands by looking inside this file. + + - "create GROUP_NAME" - creates a new security group. + + - "del GROUP_NAME" - deletes a new security group. + +Each security group's subdirectory contains 2 subdirectories: initiators +and luns. + +Each "initiators" subdirectory contains list of added to this groups +initiator as well as as well as file "mgmt". This file has the following +commands, which you can send to it, for instance, using "echo" shell +command. You can always get a small help about supported commands by +looking inside this file. + + - "add INITIATOR_NAME" - adds initiator with name INITIATOR_NAME to the + group. + + - "del INITIATOR_NAME" - deletes initiator with name INITIATOR_NAME + from the group. + + - "move INITIATOR_NAME DEST_GROUP_NAME" moves initiator with name + INITIATOR_NAME from the current group to group with name + DEST_GROUP_NAME. + + - "clear" - deletes all initiators from this group. + +For "add" and "del" commands INITIATOR_NAME can be a simple DOS-type +patterns, containing '*' and '?' symbols. '*' means match all any +symbols, '?' means match only any single symbol. For instance, +"blah.xxx" will match "bl?h.*". + +Each "luns" subdirectory contains the list of already defined LUNs for +this group as well as file "mgmt". Content of this file as well as list +of available in it commands is fully identical to the "luns" +subdirectory of the target-oriented access control. + +Examples: + + - echo "create INI" >/sys/kernel/scst_tgt/targets/iscsi/iqn.2006-10.net.vlnb:tgt1/ini_groups/mgmt - + creates security group INI for target iqn.2006-10.net.vlnb:tgt1. + + - echo "add 2:0:1:0 11" >/sys/kernel/scst_tgt/targets/iscsi/iqn.2006-10.net.vlnb:tgt1/ini_groups/INI/luns/mgmt - + adds a pass-through device sitting on host 2, channel 0, ID 1, LUN 0 + to group with name INI as LUN 11. + + - echo "add disk1 0" >/sys/kernel/scst_tgt/targets/iscsi/iqn.2006-10.net.vlnb:tgt1/ini_groups/INI/luns/mgmt - + adds a virtual disk with name disk1 to group with name INI as LUN 0. + + - echo "add 21:*:e0:?b:83:*" >/sys/kernel/scst_tgt/targets/21:00:00:a0:8c:54:52:12/ini_groups/INI/initiators/mgmt - + adds a pattern to group with name INI to Fibre Channel target with + WWN 21:00:00:a0:8c:54:52:12, which matches WWNs of Fibre Channel + initiator ports. + +Consider you need to have an iSCSI target with name +"iqn.2007-05.com.example:storage.disk1.sys1.xyz", which should export +virtual device "dev1" with LUN 0 and virtual device "dev2" with LUN 1, +but initiator with name +"iqn.2007-05.com.example:storage.disk1.spec_ini.xyz" should see only +virtual device "dev2" read only with LUN 0. To achieve that you should +do the following commands: + +# echo "iqn.2007-05.com.example:storage.disk1.sys1.xyz" >/sys/kernel/scst_tgt/targets/iscsi/mgmt +# echo "add dev1 0" >/sys/kernel/scst_tgt/targets/iscsi/iqn.2007-05.com.example:storage.disk1.sys1.xyz/luns/mgmt +# echo "add dev2 1" >/sys/kernel/scst_tgt/targets/iscsi/iqn.2007-05.com.example:storage.disk1.sys1.xyz/luns/mgmt +# echo "create SPEC_INI" >/sys/kernel/scst_tgt/targets/iscsi/iqn.2007-05.com.example:storage.disk1.sys1.xyz/ini_groups/mgmt +# echo "add dev2 0 read_only=1" \ + >/sys/kernel/scst_tgt/targets/iscsi/iqn.2007-05.com.example:storage.disk1.sys1.xyz/ini_groups/SPEC_INI/luns/mgmt +# echo "iqn.2007-05.com.example:storage.disk1.spec_ini.xyz" \ + >/sys/kernel/scst_tgt/targets/iscsi/iqn.2007-05.com.example:storage.disk1.sys1.xyz/ini_groups/SPEC_INI/initiators/mgmt + +For Fibre Channel or SAS in the above example you should use target's +and initiator ports WWNs instead of iSCSI names. + +It is highly recommended to use scstadmin utility instead of described +in this section low level interface. + +IMPORTANT +========= + +There must be LUN 0 in each set of LUNs, i.e. LUs numeration must not +start from, e.g., 1. Otherwise you will see no devices on remote +initiators and SCST core will write into the kernel log message: "tgt_dev +for LUN 0 not found, command to unexisting LU?" + +IMPORTANT +========= + +All the access control must be fully configured BEFORE the corresponding +target is enabled! When you enable a target, it will immediately start +accepting new connections, hence creating new sessions, and those new +sessions will be assigned to security groups according to the +*currently* configured access control settings. For instance, to +the default target's set of LUNs, instead of "HOST004" group as you may +need, because "HOST004" doesn't exist yet. So, one must configure all +the security groups before new connections from the initiators are +created, i.e. before the target enabled. + +VDISK device handler +-------------------- + +VDISK has 4 built-in dev handlers: vdisk_fileio, vdisk_blockio, +vdisk_nullio and vcdrom. Roots of their sysfs interface are +/sys/kernel/scst_tgt/handlers/handler_name, e.g. for vdisk_fileio: +/sys/kernel/scst_tgt/handlers/vdisk_fileio. Each root has the following +entries: + + - None, one or more links to devices with name equal to names + of the corresponding devices. + + - trace_level - allows to enable and disable various tracing + facilities. See content of this file for help how to use it. + + - mgmt - main management entry, which allows to add/delete VDISK + devices with the corresponding type. + +The "mgmt" file has the following commands, which you can send to it, +for instance, using "echo" shell command. You can always get a small +help about supported commands by looking inside this file. "Parameters" +are one or more param_name=value pairs separated by ';'. + + - echo "add_device device_name [parameters]" - adds a virtual device + with name device_name and specified parameters (see below) + + - echo "del_device device_name" - deletes a virtual device with name + device_name. + +Handler vdisk_fileio provides FILEIO mode to create virtual devices. +This mode uses as backend files and accesses to them using regular +read()/write() file calls. This allows to use full power of Linux page +cache. The following parameters possible for vdisk_fileio: + + - filename - specifies path and file name of the backend file. The path + must be absolute. + + - blocksize - specifies block size used by this virtual device. The + block size must be power of 2 and >= 512 bytes. Default is 512. + + - write_through - disables write back caching. Note, this option + has sense only if you also *manually* disable write-back cache in + *all* your backstorage devices and make sure it's actually disabled, + since many devices are known to lie about this mode to get better + benchmark results. Default is 0. + + - read_only - read only. Default is 0. + + - o_direct - disables both read and write caching. This mode isn't + currently fully implemented, you should use user space fileio_tgt + program in O_DIRECT mode instead (see below). + + - nv_cache - enables "non-volatile cache" mode. In this mode it is + assumed that the target has a GOOD UPS with ability to cleanly + shutdown target in case of power failure and it is software/hardware + bugs free, i.e. all data from the target's cache are guaranteed + sooner or later to go to the media. Hence all data synchronization + with media operations, like SYNCHRONIZE_CACHE, are ignored in order + to bring more performance. Also in this mode target reports to + initiators that the corresponding device has write-through cache to + disable all write-back cache workarounds used by initiators. Use with + extreme caution, since in this mode after a crash of the target + journaled file systems don't guarantee the consistency after journal + recovery, therefore manual fsck MUST be ran. Note, that since usually + the journal barrier protection (see "IMPORTANT" note below) turned + off, enabling NV_CACHE could change nothing from data protection + point of view, since no data synchronization with media operations + will go from the initiator. This option overrides "write_through" + option. Disabled by default. + + - removable - with this flag set the device is reported to remote + initiators as removable. + +Handler vdisk_blockio provides BLOCKIO mode to create virtual devices. +This mode performs direct block I/O with a block device, bypassing the +page cache for all operations. This mode works ideally with high-end +storage HBAs and for applications that either do not need caching +between application and disk or need the large block throughput. See +below for more info. + +The following parameters possible for vdisk_blockio: filename, +blocksize, read_only, removable. See vdisk_fileio above for description +of those parameters. + +Handler vdisk_nullio provides NULLIO mode to create virtual devices. In +this mode no real I/O is done, but success returned to initiators. +Intended to be used for performance measurements at the same way as +"*_perf" handlers. The following parameters possible for vdisk_nullio: +blocksize, read_only, removable. See vdisk_fileio above for description +of those parameters. + +Handler vcdrom allows emulation of a virtual CDROM device using an ISO +file as backend. It doesn't have any parameters. + +For example: + +echo "add_device disk1 filename=/disk1; blocksize=4096; nv_cache=1" >/sys/kernel/scst_tgt/handlers/vdisk_fileio/mgmt + +will create a FILEIO virtual device disk1 with backend file /disk1 +with block size 4K and NV_CACHE enabled. + +Each vdisk_fileio's device has the following attributes in +/sys/kernel/scst_tgt/devices/device_name: + + - filename - contains path and file name of the backend file. + + - blocksize - contains block size used by this virtual device. + + - write_through - contains status of write back caching of this virtual + device. + + - read_only - contains read only status of this virtual device. + + - o_direct - contains O_DIRECT status of this virtual device. + + - nv_cache - contains NV_CACHE status of this virtual device. + + - removable - contains removable status of this virtual device. + + - size_mb - contains size of this virtual device in MB. + + - t10_dev_id - contains and allows to set T10 vendor specific + identifier for Device Identification VPD page (0x83) of INQUIRY data. + By default VDISK handler always generates t10_dev_id for every new + created device at creation time based on the device name and + scst_vdisk_ID scst_vdisk.ko module parameter (see below). + + - usn - contains the virtual device's serial number of INQUIRY data. It + is created at the device creation time based on the device name and + scst_vdisk_ID scst_vdisk.ko module parameter (see below). + + - type - contains SCSI type of this virtual device. + + - resync_size - write only attribute, which makes vdisk_fileio to + rescan size of the backend file. It is useful if you changed it, for + instance, if you resized it. + +For example: + +/sys/kernel/scst_tgt/devices/disk1 +|-- blocksize +|-- exported +| |-- export0 -> ../../../targets/iscsi/iqn.2006-10.net.vlnb:tgt/luns/0 +| |-- export1 -> ../../../targets/iscsi/iqn.2006-10.net.vlnb:tgt/ini_groups/INI/luns/0 +| |-- export2 -> ../../../targets/iscsi/iqn.2006-10.net.vlnb:tgt1/luns/0 +| |-- export3 -> ../../../targets/iscsi/iqn.2006-10.net.vlnb:tgt1/ini_groups/INI1/luns/0 +| |-- export4 -> ../../../targets/iscsi/iqn.2006-10.net.vlnb:tgt1/ini_groups/INI2/luns/0 +|-- filename +|-- handler -> ../../handlers/vdisk_fileio +|-- nv_cache +|-- o_direct +|-- read_only +|-- removable +|-- resync_size +|-- size_mb +|-- t10_dev_id +|-- threads_num +|-- threads_pool_type +|-- type +|-- usn +`-- write_through + +Each vdisk_blockio's device has the following attributes in +/sys/kernel/scst_tgt/devices/device_name: blocksize, filename, +read_only, removable, resync_size, size_mb, t10_dev_id, threads_num, +threads_pool_type, type, usn. See above description of those parameters. + +Each vdisk_nullio's device has the following attributes in +/sys/kernel/scst_tgt/devices/device_name: blocksize, read_only, +removable, size_mb, t10_dev_id, threads_num, threads_pool_type, type, +usn. See above description of those parameters. + +Each vcdrom's device has the following attributes in +/sys/kernel/scst_tgt/devices/device_name: filename, size_mb, +t10_dev_id, threads_num, threads_pool_type, type, usn. See above +description of those parameters. Exception is filename attribute. For +vcdrom it is writable. Writing to it allows to virtually insert or +change virtual CD media in the virtual CDROM device. For example: + + - echo "/image.iso" >/sys/kernel/scst_tgt/devices/cdrom/filename - will + insert file /image.iso as virtual media to the virtual CDROM cdrom. + + - echo "" >/sys/kernel/scst_tgt/devices/cdrom/filename - will remove + "media" from the virtual CDROM cdrom. + +Additionally to the sysfs interface VDISK handler has module parameter +"num_threads", which specifies count of I/O threads for each VDISK's +device. If you have a workload, which tends to produce rather random +accesses (e.g. DB-like), you should increase this count to a bigger +value, like 32. If you have a rather sequential workload, you should +decrease it to a lower value, like number of CPUs on the target or even +1. Due to some limitations of Linux I/O subsystem, increasing number of +I/O threads too much leads to sequential performance drop, especially +with deadline scheduler, so decreasing it can improve sequential +performance. The default provides a good compromise between random and +sequential accesses. + +You shouldn't be afraid to have too many VDISK I/O threads if you have +many VDISK devices. Kernel threads consume very little amount of +resources (several KBs) and only necessary threads will be used by SCST, +so the threads will not trash your system. + +CAUTION: If you partitioned/formatted your device with block size X, *NEVER* +======== ever try to export and then mount it (even accidentally) with another + block size. Otherwise you can *instantly* damage it pretty + badly as well as all your data on it. Messages on initiator + like: "attempt to access beyond end of device" is the sign of + such damage. + + Moreover, if you want to compare how well different block sizes + work for you, you **MUST** EVERY TIME AFTER CHANGING BLOCK SIZE + **COMPLETELY** **WIPE OFF** ALL THE DATA FROM THE DEVICE. In + other words, THE **WHOLE** DEVICE **MUST** HAVE ONLY **ZEROS** + AS THE DATA AFTER YOU SWITCH TO NEW BLOCK SIZE. Switching block + sizes isn't like switching between FILEIO and BLOCKIO, after + changing block size all previously written with another block + size data MUST BE ERASED. Otherwise you will have a full set of + very weird behaviors, because blocks addressing will be + changed, but initiators in most cases will not have a + possibility to detect that old addresses written on the device + in, e.g., partition table, don't refer anymore to what they are + intended to refer. + +IMPORTANT: Some disk and partition table management utilities don't support +========= block sizes >512 bytes, therefore make sure that your favorite one + supports it. Currently only cfdisk is known to work only with + 512 bytes blocks, other utilities like fdisk on Linux or + standard disk manager on Windows are proved to work well with + non-512 bytes blocks. Note, if you export a disk file or + device with some block size, different from one, with which + it was already partitioned, you could get various weird + things like utilities hang up or other unexpected behavior. + Hence, to be sure, zero the exported file or device before + the first access to it from the remote initiator with another + block size. On Window initiator make sure you "Set Signature" + in the disk manager on the imported from the target drive + before doing any other partitioning on it. After you + successfully mounted a file system over non-512 bytes block + size device, the block size stops matter, any program will + work with files on such file system. + +Caching +------- + +By default for performance reasons VDISK FILEIO devices use write back +caching policy. This is generally safe for modern applications who +prepared to work in the write back caching environments, so know when to +flush cache to keep their data consistent and minimize damage caused in +case of power/hardware/software failures by lost in the cache data. + +For instance, journaled file systems flush cache on each meta data +update, so they survive power/hardware/software failures pretty well. +Note, Linux IO subsystem guarantees it work reliably only using data +protection barriers, which, for instance, for Ext3 turned off by default +(see http://lwn.net/Articles/283161). Some info about barriers from the +XFS point of view could be found at +http://oss.sgi.com/projects/xfs/faq.html#wcache. On Linux initiators for +Ext3 and ReiserFS file systems the barrier protection could be turned on +using "barrier=1" and "barrier=flush" mount options correspondingly. You +can check if the barriers turn on or off by looking in /proc/mounts. +Windows and, AFAIK, other UNIX'es don't need any special explicit +options and do necessary barrier actions on write-back caching devices +by default. + +But even in case of journaled file systems your unsaved cached data will +still be lost in case of power/hardware/software failures, so you may +need to supply your target server with a good UPS with possibility to +gracefully shutdown your target on power shortage or disable write back +caching using WRITE_THROUGH flag. Note, on some real-life workloads +write through caching might perform better, than write back one with the +barrier protection turned on. Also note that without barriers enabled +(i.e. by default) Linux doesn't provide a guarantee that after +sync()/fsync() all written data really hit permanent storage. They can +be stored in the cache of your backstorage devices and, hence, lost on a +power failure event. Thus, ever with write-through cache mode, you still +either need to enable barriers on your backend file system on the target +(for devices this is, indeed, impossible), or need a good UPS to protect +yourself from not committed data loss. + +To limit this data loss you can use files in /proc/sys/vm to limit +amount of unflushed data in the system cache. + +BLOCKIO VDISK mode +------------------ + +This module works best for these types of scenarios: + +1) Data that are not aligned to 4K sector boundaries and <4K block sizes +are used, which is normally found in virtualization environments where +operating systems start partitions on odd sectors (Windows and it's +sector 63). + +2) Large block data transfers normally found in database loads/dumps and +streaming media. + +3) Advanced relational database systems that perform their own caching +which prefer or demand direct IO access and, because of the nature of +their data access, can actually see worse performance with +non-discriminate caching. + +4) Multiple layers of targets were the secondary and above layers need +to have a consistent view of the primary targets in order to preserve +data integrity which a page cache backed IO type might not provide +reliably. + +Also it has an advantage over FILEIO that it doesn't copy data between +the system cache and the commands data buffers, so it saves a +considerable amount of CPU power and memory bandwidth. + +IMPORTANT: Since data in BLOCKIO and FILEIO modes are not consistent between +========= them, if you try to use a device in both those modes simultaneously, + you will almost instantly corrupt your data on that device. + +Pass-through mode +----------------- + +In the pass-through mode (i.e. using the pass-through device handlers +scst_disk, scst_tape, etc) SCSI commands, coming from remote initiators, +are passed to local SCSI devices on target as is, without any +modifications. + +In the SYSFS interface all real SCSI devices are listed in +/sys/kernel/scst_tgt/devices in form host:channel:id:lun numbers, for +instance 1:0:0:0. The recommended way to match those numbers to your +devices is use of lsscsi utility. + +When a pass-through dev handler is loaded it assigns itself to all +existing SCSI devices of its SCSI type. If you later want to unassign some +SCSI device from it or assign it to another dev handler you can use the +following interface. + +Each pass-through dev handler has in its root subdirectory +/sys/kernel/scst_tgt/handlers/handler_name, e.g. +/sys/kernel/scst_tgt/handlers/dev_disk, "mgmt" file. It allows the +following commands. They can be sent to it using, e.g., echo command. + + - "assign" - this command assigns SCSI device with +host:channel:id:lun numbers to this dev handler. + +echo "assign 1:0:0:0" >mgmt + +will assign SCSI device 1:0:0:0 to this dev handler. + + - "unassign" - this command unassigns SCSI device with +host:channel:id:lun numbers from this dev handler. + +As usually, on read the "mgmt" file returns small help about available +commands. + +As any other hardware, the local SCSI hardware can not handle commands +with amount of data and/or segments count in scatter-gather array bigger +some values. Therefore, when using the pass-through mode you should note +that values for maximum number of segments and maximum amount of +transferred data for each SCSI command on devices on initiators can not +be bigger, than corresponding values of the corresponding SCSI devices +on the target. Otherwise you will see symptoms like small transfers work +well, but large ones stall and messages like: "Unable to complete +command due to SG IO count limitation" are printed in the kernel logs. + +You can't control from the user space limit of the scatter-gather +segments, but for block devices usually it is sufficient if you set on +the initiators /sys/block/DEVICE_NAME/queue/max_sectors_kb in the same +or lower value as in /sys/block/DEVICE_NAME/queue/max_hw_sectors_kb for +the corresponding devices on the target. + +For not-block devices SCSI commands are usually generated directly by +applications, so, if you experience large transfers stalls, you should +check documentation for your application how to limit the transfer +sizes. + +Another way to solve this issue is to build SG entries with more than 1 +page each. See the following patch as an example: +http://scst.sourceforge.net/sgv_big_order_alloc.diff + +User space mode using scst_user dev handler +------------------------------------------- + +User space program fileio_tgt uses interface of scst_user dev handler +and allows to see how it works in various modes. Fileio_tgt provides +mostly the same functionality as scst_vdisk handler with the most +noticeable difference that it supports O_DIRECT mode. O_DIRECT mode is +basically the same as BLOCKIO, but also supports files, so for some +loads it could be significantly faster, than the regular FILEIO access. +All the words about BLOCKIO from above apply to O_DIRECT as well. See +fileio_tgt's README file for more details. + +Performance +----------- + +SCST from the very beginning has been designed and implemented to +provide the best possible performance. Since there is no "one fit all" +the best performance configuration for different setups and loads, SCST +provides extensive set of settings to allow to tune it for the best +performance in each particular case. You don't have to necessary use +those settings. If you don't, SCST will do very good job to autotune for +you, so the resulting performance will, in average, be better +(sometimes, much better) than with other SCSI targets. But in some cases +you can by manual tuning improve it even more. + +Before doing any performance measurements note that performance results +are very much dependent from your type of load, so it is crucial that +you choose access mode (FILEIO, BLOCKIO, O_DIRECT, pass-through), which +suits your needs the best. + +In order to get the maximum performance you should: + +1. For SCST: + + - Disable in Makefile CONFIG_SCST_STRICT_SERIALIZING, CONFIG_SCST_EXTRACHECKS, + CONFIG_SCST_TRACING, CONFIG_SCST_DEBUG*, CONFIG_SCST_STRICT_SECURITY + + - For pass-through devices enable + CONFIG_SCST_ALLOW_PASSTHROUGH_IO_SUBMIT_IN_SIRQ. + +2. For target drivers: + + - Disable in Makefiles CONFIG_SCST_EXTRACHECKS, CONFIG_SCST_TRACING, + CONFIG_SCST_DEBUG* + +3. For device handlers, including VDISK: + + - Disable in Makefile CONFIG_SCST_TRACING and CONFIG_SCST_DEBUG. + +IMPORTANT: Some of the above compilation options in the SCST SVN enabled +========= by default, i.e. the development version of SCST is optimized + for development and bug hunting, not for performance. For it + you can set the above options, except + CONFIG_SCST_ALLOW_PASSTHROUGH_IO_SUBMIT_IN_SIRQ, in the + needed values by command "make debug2perf" performed in + trunk/. + +4. Make sure you have io_grouping_type option set correctly, especially +in the following cases: + + - Several initiators share your target's backstorage. It can be a + shared LU using some cluster FS, like VMFS, as well as can be + different LUs located on the same backstorage (RAID array). For + instance, if you have 3 initiators and each of them using its own + dedicated FILEIO device file from the same RAID-6 array on the + target. + + In this case for the best performance you should have + io_grouping_type option set in value "never" in all the LUNs' targets + and security groups. + + - Your initiator connected to your target in MPIO mode. In this case for + the best performance you should: + + * Either connect all the sessions from the initiator to a single + target or security group and have io_grouping_type option set in + value "this_group_only" in the target or security group, + + * Or, if it isn't possible to connect all the sessions from the + initiator to a single target or security group, assign the same + numeric io_grouping_type value for each target/security group this + initiator connected to. The exact value itself doesn't matter, + important only that all the targets/security groups use the same + value. + +Don't forget, io_grouping_type makes sense only if you use CFQ I/O +scheduler on the target and for devices with threads_num >= 0 and, if +threads_num > 0, with threads_pool_type "per_initiator". + +You can check if in your setup io_grouping_type set correctly as well as +if the "auto" io_grouping_type value works for you by tests like the +following: + + - For not MPIO case you can run single thread sequential reading, e.g. + using buffered dd, from one initiator, then run the same single + thread sequential reading from the second initiator in parallel. If + io_grouping_type is set correctly the aggregate throughput measured + on the target should only slightly decrease as well as all initiators + should have nearly equal share of it. If io_grouping_type is not set + correctly, the aggregate throughput and/or throughput on any + initiator will decrease significantly, in 2 times or even more. For + instance, you have 80MB/s single thread sequential reading from the + target on any initiator. When then both initiators are reading in + parallel you should see on the target aggregate throughput something + like 70-75MB/s with correct io_grouping_type and something like + 35-40MB/s or 8-10MB/s on any initiator with incorrect. + + - For the MPIO case it's quite easier. With incorrect io_grouping_type + you simply won't see performance increase from adding the second + session (assuming your hardware is capable to transfer data through + both sessions in parallel), or can even see a performance decrease. + +5. If you are going to use your target in an VM environment, for +instance as a shared storage with VMware, make sure all your VMs +connected to the target via *separate* sessions. For instance, for iSCSI +it means that each VM has own connection to the target, not all VMs +connected using a single connection. You can check it using SCST sysfs +interface. For other transports you should use available facilities, +like NPIV for Fibre Channel, to make separate sessions for each VM. If +you miss it, you can greatly loose performance of parallel access to +your target from different VMs. This isn't related to the case if your +VMs are using the same shared storage, like with VMFS, for instance. In +this case all your VM hosts will be connected to the target via separate +sessions, which is enough. + +6. For other target and initiator software parts: + + - Make sure you applied on your kernel all available SCST patches. + If for your kernel version this patch doesn't exist, it is strongly + recommended to upgrade your kernel to version, for which this patch + exists. + + - Don't enable debug/hacking features in the kernel, i.e. use them as + they are by default. + + - The default kernel read-ahead and queuing settings are optimized + for locally attached disks, therefore they are not optimal if they + attached remotely (SCSI target case), which sometimes could lead to + unexpectedly low throughput. You should increase read-ahead size to at + least 512KB or even more on all initiators and the target. + + You should also limit on all initiators maximum amount of sectors per + SCSI command. This tuning is also recommended on targets with large + read-ahead values. To do it on Linux, run: + + echo “64” > /sys/block/sdX/queue/max_sectors_kb + + where specify instead of X your imported from target device letter, + like 'b', i.e. sdb. + + To increase read-ahead size on Linux, run: + + blockdev --setra N /dev/sdX + + where N is a read-ahead number in 512-byte sectors and X is a device + letter like above. + + Note: you need to set read-ahead setting for device sdX again after + you changed the maximum amount of sectors per SCSI command for that + device. + + Note2: you need to restart SCST after you changed read-ahead settings + on the target. + + - You may need to increase amount of requests that OS on initiator + sends to the target device. To do it on Linux initiators, run + + echo “64” > /sys/block/sdX/queue/nr_requests + + where X is a device letter like above. + + You may also experiment with other parameters in /sys/block/sdX + directory, they also affect performance. If you find the best values, + please share them with us. + + - On the target use CFQ IO scheduler. In most cases it has performance + advantage over other IO schedulers, sometimes huge (2+ times + aggregate throughput increase). + + - It is recommended to turn the kernel preemption off, i.e. set + the kernel preemption model to "No Forced Preemption (Server)". + + - Looks like XFS is the best filesystem on the target to store device + files, because it allows considerably better linear write throughput, + than ext3. + +7. For hardware on target. + + - Make sure that your target hardware (e.g. target FC or network card) + and underlaying IO hardware (e.g. IO card, like SATA, SCSI or RAID to + which your disks connected) don't share the same PCI bus. You can + check it using lspci utility. They have to work in parallel, so it + will be better if they don't compete for the bus. The problem is not + only in the bandwidth, which they have to share, but also in the + interaction between cards during that competition. This is very + important, because in some cases if target and backend storage + controllers share the same PCI bus, it could lead up to 5-10 times + less performance, than expected. Moreover, some motherboard (by + Supermicro, particularly) have serious stability issues if there are + several high speed devices on the same bus working in parallel. If + you have no choice, but PCI bus sharing, set in the BIOS PCI latency + as low as possible. + +8. If you use VDISK IO module in FILEIO mode, NV_CACHE option will +provide you the best performance. But using it make sure you use a good +UPS with ability to shutdown the target on the power failure. + +Baseline performance numbers you can find in those measurements: +http://lkml.org/lkml/2009/3/30/283. + +IMPORTANT: If you use on initiator some versions of Windows (at least W2K) +========= you can't get good write performance for VDISK FILEIO devices with + default 512 bytes block sizes. You could get about 10% of the + expected one. This is because of the partition alignment, which + is (simplifying) incompatible with how Linux page cache + works, so for each write the corresponding block must be read + first. Use 4096 bytes block sizes for VDISK devices and you + will have the expected write performance. Actually, any OS on + initiators, not only Windows, will benefit from block size + max(PAGE_SIZE, BLOCK_SIZE_ON_UNDERLYING_FS), where PAGE_SIZE + is the page size, BLOCK_SIZE_ON_UNDERLYING_FS is block size + on the underlying FS, on which the device file located, or 0, + if a device node is used. Both values are from the target. + See also important notes about setting block sizes >512 bytes + for VDISK FILEIO devices above. + +9. In some cases, for instance working with SSD devices, which consume 100% +of a single CPU load for data transfers in their internal threads, to +maximize IOPS it can be needed to assign for those threads dedicated +CPUs using Linux CPU affinity facilities. No IRQ processing should be +done on those CPUs. Check that using /proc/interrupts. See taskset +command and Documentation/IRQ-affinity.txt in your kernel's source tree +for how to assign IRQ affinity to tasks and IRQs. + +The reason for that is that processing of coming commands in SIRQ +context might be done on the same CPUs as SSD devices' threads doing data +transfers. As the result, those threads won't receive all the processing +power of those CPUs and perform worse. + +Work if target's backstorage or link is too slow +------------------------------------------------ + +Under high I/O load, when your target's backstorage gets overloaded, or +working over a slow link between initiator and target, when the link +can't serve all the queued commands on time, you can experience I/O +stalls or see in the kernel log abort or reset messages. + +At first, consider the case of too slow target's backstorage. On some +seek intensive workloads even fast disks or RAIDs, which able to serve +continuous data stream on 500+ MB/s speed, can be as slow as 0.3 MB/s. +Another possible cause for that can be MD/LVM/RAID on your target as in +http://lkml.org/lkml/2008/2/27/96 (check the whole thread as well). + +Thus, in such situations simply processing of one or more commands takes +too long time, hence initiator decides that they are stuck on the target +and tries to recover. Particularly, it is known that the default amount +of simultaneously queued commands (48) is sometimes too high if you do +intensive writes from VMware on a target disk, which uses LVM in the +snapshot mode. In this case value like 16 or even 8-10 depending of your +backstorage speed could be more appropriate. + +Unfortunately, currently SCST lacks dynamic I/O flow control, when the +queue depth on the target is dynamically decreased/increased based on +how slow/fast the backstorage speed comparing to the target link. So, +there are 6 possible actions, which you can do to workaround or fix this +issue in this case: + +1. Ignore incoming task management (TM) commands. It's fine if there are +not too many of them, so average performance isn't hurt and the +corresponding device isn't getting put offline, i.e. if the backstorage +isn't too slow. + +2. Decrease /sys/block/sdX/device/queue_depth on the initiator in case +if it's Linux (see below how) or/and SCST_MAX_TGT_DEV_COMMANDS constant +in scst_priv.h file until you stop seeing incoming TM commands. +ISCSI-SCST driver also has its own iSCSI specific parameter for that, +see its README file. + +To decrease device queue depth on Linux initiators you can run command: + +# echo Y >/sys/block/sdX/device/queue_depth + +where Y is the new number of simultaneously queued commands, X - your +imported device letter, like 'a' for sda device. There are no special +limitations for Y value, it can be any value from 1 to possible maximum +(usually, 32), so start from dividing the current value on 2, i.e. set +16, if /sys/block/sdX/device/queue_depth contains 32. + +3. Increase the corresponding timeout on the initiator. For Linux it is +located in +/sys/devices/platform/host*/session*/target*:0:0/*:0:0:1/timeout. It can +be done automatically by an udev rule. For instance, the following +rule will increase it to 300 seconds: + +SUBSYSTEM=="scsi", KERNEL=="[0-9]*:[0-9]*", ACTION=="add", ATTR{type}=="0|7|14", ATTR{timeout}="300" + +By default, this timeout is 30 or 60 seconds, depending on your distribution. + +4. Try to avoid such seek intensive workloads. + +5. Increase speed of the target's backstorage. + +6. Implement in SCST dynamic I/O flow control. This will be an ultimate +solution. See "Dynamic I/O flow control" section on +http://scst.sourceforge.net/contributing.html page for possible +implementation idea. + +Next, consider the case of too slow link between initiator and target, +when the initiator tries to simultaneously push N commands to the target +over it. In this case time to serve those commands, i.e. send or receive +data for them over the link, can be more, than timeout for any single +command, hence one or more commands in the tail of the queue can not be +served on time less than the timeout, so the initiator will decide that +they are stuck on the target and will try to recover. + +To workaround/fix this issue in this case you can use ways 1, 2, 3, 6 +above or (7): increase speed of the link between target and initiator. +But for some initiators implementations for WRITE commands there might +be cases when target has no way to detect the issue, so dynamic I/O flow +control will not be able to help. In those cases you could also need on +the initiator(s) to either decrease the queue depth (way 2), or increase +the corresponding timeout (way 3). + +Note, that logged messages about QUEUE_FULL status are quite different +by nature. This is a normal work, just SCSI flow control in action. +Simply don't enable "mgmt_minor" logging level, or, alternatively, if +you are confident in the worst case performance of your back-end storage +or initiator-target link, you can increase SCST_MAX_TGT_DEV_COMMANDS in +scst_priv.h to 64. Usually initiators don't try to push more commands on +the target. + +Credits +------- + +Thanks to: + + * Mark Buechler <mark.buechler@xxxxxxxxx> for a lot of useful + suggestions, bug reports and help in debugging. + + * Ming Zhang <mingz@xxxxxxxxxxx> for fixes and comments. + + * Nathaniel Clark <nate@xxxxxxxxxx> for fixes and comments. + + * Calvin Morrow <calvin.morrow@xxxxxxxxxxx> for testing and useful + suggestions. + + * Hu Gang <hugang@xxxxxxxxxxxx> for the original version of the + LSI target driver. + + * Erik Habbinga <erikhabbinga@xxxxxxxxxxxxxxxx> for fixes and support + of the LSI target driver. + + * Ross S. W. Walker <rswwalker@xxxxxxxxxxx> for the original block IO + code and Vu Pham <huongvp@xxxxxxxxx> who updated it for the VDISK dev + handler. + + * Michael G. Byrnes <michael.byrnes@xxxxxx> for fixes. + + * Alessandro Premoli <a.premoli@xxxxxxxxx> for fixes + + * Nathan Bullock <nbullock@xxxxxxxxxxxxxx> for fixes. + + * Terry Greeniaus <tgreeniaus@xxxxxxxxxxxxxx> for fixes. + + * Krzysztof Blaszkowski <kb@xxxxxxxxxxxxxxx> for many fixes and bug reports. + + * Jianxi Chen <pacers@xxxxxxxxxxxxxxxxxxxxx> for fixing problem with + devices >2TB in size + + * Bart Van Assche <bart.vanassche@xxxxxxxxx> for a lot of help + + * Daniel Debonzi <debonzi@xxxxxxxxxxxxxxxxxx> for a big part of SCST sysfs tree + implementation + +Vladislav Bolkhovitin <vst@xxxxxxxx>, http://scst.sourceforge.net -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
