linux-iscsi 4.0.1.11 is now available on SourceForge
<http://sf.net/projects/linux-iscsi/>
 
Notes:
=====
Release "linux-iscsi-4.0.1.11" consists of following three files:
 
1.linux-iscsi-kernel-4.0.1.11.patch - iSCSI kernel patch for linux kernel version 2.6.10
 
2.linux-iscsi-user-4.0.1.11.tgz - Contains iSCSI daemon and utilities.
 
3.linux-iscsi-4.0.1.11.tgz - Complete tar-ball.
 
Change Log:
=========
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             Change the session_drop_time and
             session_established_time attributes and
             correspondingly change iscsi-ls
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             iscsi-ls.patch
             Adds the iscsi-ls utility
PATCH - Naveen Burmi [naveenb@cisco.com]
             minor Makefile cleanups
PATCH - Krishna Murthy  N.C [krmurthy@cisco.com]
             README changes for 4.0.1.11
PATCH - Naveen Burmi [naveenb@cisco.com]
             Fixed the sourceforge bug 1060754 - Error while
             doing the make install of the iscsi driver
PATCH - Naveen Burmi [naveenb@cisco.com]
             check for the existence of /dev/iscsictl, before
             removing it while uninstalling the driver
PATCH - Mike Christie [mikenc@us.ibm.com]
             add crc drop session fix per review comments,
             rm my removal of login return code,
             and comment out __scsi_print_sense usage until
             2.6.11
PATCH - Mike Christie [mikenc@us.ibm.com]
             cleanup from iscsi-io removal
PATCH - Mike Christie [mikenc@us.ibm.com]
             dont kmalloc tfm when IO is outstanding
PATCH - Mike Christie [mikenc@us.ibm.com]
             use common IO return values
PATCH - Mike Christie [mikenc@us.ibm.com]
             add whitespaces between vars and code
PATCH - Mike Christie [mikenc@us.ibm.com]
             make common iscsi def file for mainline
PATCH - Mike Christie [mikenc@us.ibm.com]
             add host info to login msgs
PATCH - Mike Christie [mikenc@us.ibm.com]
             test for portal group tag in find session
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm defferred sense.
             TODO - doc this new README
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm old return value not used
PATCH - Mike Christie [mikenc@us.ibm.com]
             dont send requests when logout is in
             progresss.
PATCH - Mike Christie [mikenc@us.ibm.com]
             dont grab and take task lock an
             extra time
PATCH - Mike Christie [mikenc@us.ibm.com]
             some initiator cleanup
PATCH - Mike Christie [mikenc@us.ibm.com]
             use tabs in iscsi-protocol.h
PATCH - Mike Christie [mikenc@us.ibm.com]
             rename sg
             TODO use sg_init_one (see linux-scsi
             patch)
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm iscsi-io header
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm extra fbit test
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm unused tx_iov
PATCH - Mike Christie [mikenc@us.ibm.com]
             xmit pdu cleanup
PATCH - Mike Christie [mikenc@us.ibm.com]
             rename add/pop task to better descibe what they do
PATCH - Mike Christie [mikenc@us.ibm.com]
             trivial create session cleanup
PATCH - Mike Christie [mikenc@us.ibm.com]
             fix printing of wrong var for digest errors
PATCH - Mike Christie [mikenc@us.ibm.com]
             cleanup from initial recv cleanup and
             add fix for non-sg data recvs (data offset should not
             be set in SETSG)
PATCH - Mike Christie [mikenc@us.ibm.com]
             cleanup some coding style issues
             in login code (rm extra brackets)
PATCH - Mike Christie [mikenc@us.ibm.com]
             mv common status code to a function
PATCH - Mike Christie [mikenc@us.ibm.com]
             address comemnts from Krishna on read pad/digest
             fn patch
PATCH - Mike Christie [mikenc@us.ibm.com]
             use scsi sense headers
PATCH - Mike Christie [mikenc@us.ibm.com]
             break out drop data into own function
 TODO: make core recv fns use it. I think there is a
             net function to do this that will be much better to
             use
PATCH - Mike Christie [mikenc@us.ibm.com]
             add helper to read digests and padding
PATCH - Mike Christie [mikenc@us.ibm.com]
             cleanup core recv fns
PATCH - Mike Christie [mikenc@us.ibm.com]
             cleanup login_response_status
PATCH - Mike Christie [mikenc@us.ibm.com]
             Must use 2.6.10 (see README)
             include scsi_transport_iscsi with linux-iscsi.
             you need to either
             insmod driver/scsi_transport_iscsi.ko
            insmod driver/iscsi_sfnet.ko
             or if everything is installed and depmod was run just
             do modprobe iscsi_sfnet
PATCH - Mike Christie [mikenc@us.ibm.com]
             dont need found var and test in find session
PATCH - Krishna Murthy  N.C [krmurthy@cisco.com]
             some iscsi-network.c cleanup
PATCH - Mike Christie [mikenc@us.ibm.com]
             tmp revert change queue depth usage. Do not
             add back queue depth attr though as it must go
             for mainline
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             fix incorrect check in get_op_params_text_keys
PATCH - Mike Christie [mikenc@us.ibm.com]
             fixup makefile per Naveen's comments (from
             dev string patch)
PATCH - Mike Christie [mikenc@us.ibm.com]
             fix shutdown/update race
PATCH - Mike Christie [mikenc@us.ibm.com]
             If a login times out, there is no session
             so drop session does not work. And, 
             when rx_establish_session calls
             terminate_session, a logout or session drop
             never occurs.
PATCH - Mike Christie [mikenc@us.ibm.com]
             cleanup oom comment
PATCH - Mike Christie [mikenc@us.ibm.com]
             Attached patch should allow someone to restart
             iscsid. Addressed Krishna's comment in next fix.
PATCH - Mike Christie [mikenc@us.ibm.com]
             The first session drop/TMF patch needed this patch
PATCH - Mike Christie [mikenc@us.ibm.com]
             The ioctl code no longer uses the host after
             the interface is exposed so the extra handle
             is not necessary. Also a state bit is needed
             to prevent someone from accessing the shutdown
             attr before the scsi_scan_host has been called.
             Yes, that would be a rare case, and most of
             the time scsi_scan_host would be called first
             and it would get the scan_mutex - just in case
             though.
PATCH - Mike Christie [mikenc@us.ibm.com]
             You can use the extra-version in
             ISCSI_DRIVER_VERSION
             instead of BUILD_STR.
PATCH - Mike Christie [mikenc@us.ibm.com]
             mv ioctl setup code to iscsi-ioctl.c
PATCH - Mike Christie [mikenc@us.ibm.com]
             sock_getsockopt was unexported in recent kernels
             and it turns out one use in the driver was bogus as
             sock_setsockopt does not return failure when you
             try to set it too high. We may want
             to try NFS approach.
             TODO - handle connection error log msg better.
PATCH - Mike Christie [mikenc@us.ibm.com]
             the driver functions return 0 for success and some -
             Exxx valuefor error, but the ioctl used 1 for success.
             The attached patch makes iscsid use 0 for success,
             and cleans up the ioctl error code.
PATCH - Mike Christie [mikenc@us.ibm.com]
             use iscsi transport class
PATCH - Mike Christie [mikenc@us.ibm.com]
             use new queue depth attr. This will be backed
             out in later patches while we wait for scsi-misc, but
             there were too many patches built over this one.
PATCH - Mike Christie [mikenc@us.ibm.com]
              fixup and cleanup calculate_login_delay.
             session_failures usage was broken, login_delay
             did not have to be passed by ptr, use
             msleep_interruptible,
             use max_t, rm some unneded comments.
PATCH - Mike Christie [mikenc@us.ibm.com]
             only iscsi-task.c uses __iscsi_get_task.c so
             make it static and inline
PATCH - Mike Christie [mikenc@us.ibm.com]
             missing space in fn args.
PATCH - Mike Christie [mikenc@us.ibm.com]
             The attached patch removes the idd code
             and replaces it with a msleep  wait. Perf is not
             critical here, and the wait should be very short (as
             long as we are moving again).
PATCH - Mike Christie [mikenc@us.ibm.com]
             The attached patch should fix a lockup that
             occurs when a session drop removes a TMF
             task. In reinit_session, I should not
             have been resetting the TMF task's fields,
             becuase upon completion of that task, its
             state was gettting lost.
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             Remove null_callback and message_handle
             Change user_handle to session_handle
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm old !session test in allocate_itt
PATCH - Mike Christie [mikenc@us.ibm.com]
             consolidate disconnect calls in STATUS_CLASS_INITIATOR_ERR block
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm some task struct atomic ops. flags and refcount
            are protectedunder the task lock. crc bit is
            protected becuase always accessed by rx thread.
PATCH - Mike Christie [mikenc@us.ibm.com]
             session->generation no longer used
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm old includes in iscsi-task
PATCH - Mike Christie [mikenc@us.ibm.com]
             rely on scsi-ml to handle check conditions with no
             sense properly
PATCH - Mike Christie [mikenc@us.ibm.com]
             term bit can never be set while the eh is up
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             iscsi-crc.c cleanup
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             Make session->target_alias dynamic
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             remove check_iscsi_ioctld and the ioctl_size and portal_info_size from
             the ioctl structure.
PATCH - Mike Christie [mikenc@us.ibm.com]
             do not need SESSION_SCSI_EH_UP hack
PATCH - Mike Christie [mikenc@us.ibm.com]
             the only time a cmnd is pending is when it enters
             the driver. it used to be when the session drops
              too. this patch cleans up iscsi_eh_timedout
              accordingly.
PATCH - Mike Christie [mikenc@us.ibm.com]
             this patch replaces dashes for colons in some odd
             places in the docbook comments
PATCH - Mike Christie [mikenc@us.ibm.com]
             allow scsi-ml to have the final say in whether or not
             to retry a command. Also fix bug when a command 
             completes before a TMF completes and rm some
             other old tests.
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             iscsi_recvmsg cleanup
PATCH - Mike Christie [mikenc@us.ibm.com]
             cleanup file headers
PATCH - Dave Wysochanski [davidw@netapp.com]
             Cleanup returns from kernel_sendmsg(), revive
             iscsi_sendmsg() and handle signal returns
            everywhere, kill print_recv_err().
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             More cleanup of iscsi_update_session to return the
             correct values to the calling function.
PATCH - Mike Christie [mikenc@us.ibm.com]
             We have allocate_auth_buffers and
            assign_session_auth_buffers,
             so this patch just renames the latter to
            assign_auth_buffers.
PATCH - Mike Christie [mikenc@us.ibm.com]
             The last destroy_session patch was a bad diff. Need
             to also set the drop time.
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             Cleanup of session update and create functions and
            related string functions
PATCH - Mike Christie [mikenc@us.ibm.com]
             remove host_no, channel, and target_id. These can
            be found on the host if needed.
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             cleanup iscsi.h (remove old LOG macros).
PATCH - Mike Christie [mikenc@us.ibm.com]
             log_name is no longer used. The attached patch
           just removes it.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Make common code use same style as driver for
           multi-line comments.
PATCH - Mike Christie [mikenc@us.ibm.com]
             If the session is already down using
            iscsi_drop_session() will not work for termination.
PATCH - Mike Christie [mikenc@us.ibm.com]
             We could either send the mgmt task from the tx
            thread or send it from the eh thread and have the rx
            thread
             wait for all tasks to be cleaned (in other words add
             some sort of refcounting on the connection). The
             latter is overkill (but we could maybe just use one of
            the num task
             counters with some tiny modifications), so this just
             makes the tx thread send the mgmt task.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Compilation of eh cleanups patches.
                 - kill tmf and tmf_timer on iscsi-task.
                 - prealloc a task for eh.
                 - make abort task use exec task mgmt fn
PATCH - Mike Christie [mikenc@us.ibm.com]
             When we switched to kernel_*msg we didn't we
            switch to kvecs. This patch converts us.
PATCH - Mike Christie [mikenc@us.ibm.com]
             If there is no session the eh_abort/reset functions
             do not have much use. This patch uses the
             eh_timed_out
             callback to handle this in most cases. TODO - add
             sane timeout defaults and better document how they
             work now that we are not killing scsi-mls.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Add wait_for_session function. ReplacementTimeout
             is now ticking from the time we start to re-establish
             a new session instead of
             the drop time (this will be fixecd later).
            Confailtimeout also temporarly removed until timeout
            update and timer fix up is complete.
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             move auth buffer allocation out of establish_session.
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm last __KERNEL__ from shared code
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             Handle data digest errors in case of tape devices.
FIX - Mike Christie [mikenc@us.ibm.com]
             Forgot to do a "cvs remove iscsi-login.h when "rm
             iscsi-limits.h" patch was merged below.
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             Remove references to portal failover from
            documentation
PATCH - Naveen Burmi [naveenb@cisco.com]
             iscsi_make_text_pdu is moved from common code to
             userspace code.
PATCH - Mike Christie [mikenc@us.ibm.com]
             change fn protos in iscsi-auth-client.h to look like
            the driver code
PATCH - Mike Christie [mikenc@us.ibm.com]
             use get_random_bytes directly like the str*
            functions in the kernel code
PATCH - Mike Christie [mikenc@us.ibm.com]
             removes some unused code in the auth code and
             makes some code static.
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             remove portal failover from user level code.
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             remove portal failover from kernel - changes as
             suggested by Mike.
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             cleanup of fill_task_immediate_data
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             Convert printk to iscsi_err, iscsi_warn, ... and
             iscsi_host_err, iscsi_host_*
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
 f            urther cleanup in iscsi_kmap_sg and
            iscsi_kunmap_sg
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm iscsi-limits.h, and move values to iscsi.h or iscsi-
             session.h
PATCH - Mike Christie [mikenc@us.ibm.com]
             iscsi_xmit_task sends the SCSI command PDU so
            move to iscsi-xmit-pdu.c
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             remove iscsi_update_address function pointer from
            iscsi_session. Instead call function directly.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Many functions were almost using the
             Documentation doc book style.
             This patch just cleans them up.
PATCH - Mike Christie [mikenc@us.ibm.com]
 i            scsi_logout_and_disconnect() will only work in
            userspace. The kernel code has its own logout code.
            Mv this function to up to avoid confusion as to how
            it would work with the kerne lcode, and why we do
            not use it for the driver code.
 TODO: will try to merge common functionality into
             new function + fix kernel code to be able to work
             correctly, and add new function to common login
             code for both code bases to use.
PATCH - Mike Christie [mikenc@us.ibm.com]
             fixes up iscsi-login.c function comments to use the
             doc book style in the Documentation dir.
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             remove portal failover - kernel only changes
PATCH - Mike Christie [mikenc@us.ibm.com]
              make function declarations consistent when they
             return ponters.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Move functions from iscsi-initiator.c to iscsi-
             session.c that are session related to better match
              comments at top of both files.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Put header includes in correct order.
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             Use kernel_recvmsg and kernel_sendmsg instead of
             iscsi_recvmsg and iscsi_sendmsg
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             iscsi_kmap_sg now unmaps mapped pages - more
             cleanup
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             Change iscsi_strtoul to simple_strtoul in the kernel
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             iscsi_kmap_sg now unmaps mapped pages in the 'sg'
            if kmap of of any page fails - Incorporated changes
            as per Mike's comments
PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             Converts all the uint* in the kernel to u*
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             iscsi_kmap_sg now unmaps mapped pages in the 'sg'
             if kmap of any page fails.
PATCH - Mike Christie [mikenc@us.ibm.com]
             mv #include to proper place (according to linux
             coding style).
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             Added a comment mentioning the reason for
             preallocating rx_tfm and tx_tfm.
 PATCH - Smitha Narayanaswamy [smithan@cisco.com]
             Cleanup of the naming of all the externed functions
             to iscsi_
PATCH - Mike Christie [mikenc@us.ibm.com]
             Coding style cleanup and rearrange the logic to
             permit more code to be written on each line (less
             tabs) in update_sn.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Rearrange logic to permit more code to be written
             on each line (less tabbing and and less breaking up
             of lines) in iscsi_recv_reject.
PATCH - Mike Christie [mikenc@us.ibm.com]
             cleanup up old code (task->scsi_cmnd will never be
             null there) in iscsi_recv_r2t().
PATCH - Mike Christie [mikenc@us.ibm.com]
             Cleanup old some old comments and merge similar
             error messages in iscsi-recv-task-mgmt().
PATCH - Mike Christie [mikenc@us.ibm.com]
             edit locking comments at top of files (missed from
             bh patch)
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             handle data digest errors
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             use spin_lock_bh and spin_unlock_bh for task_lock.
PATCH - Krishna Murthy N.C [krmurthy@cisco.com]
             scsi_cmnd->result now contains valid values.
#1060875 - disk_command_timeout needs to be removed from
                user space Removes disk_command_timeout from
                user-level files and  DiskCommandTimeout from
                the documentation and iscsi.conf
PATCH - Mike Christie [mikenc@us.ibm.com]
             update README to reflect new required kernel
              version
PATCH - Mike Christie [mikenc@us.ibm.com]
              adds a task management task completion function
             so that the recv code does not have to figure out
             when to wake the scsi error thread.This will be
             merged with the normal task completion function
             when the tx code is cleaned up.
PATCH - Mike Christie [mikenc@us.ibm.com]
             wake up the eh thread if the session is dropped
             while the eh is running and the task is a abort or a
             reset (previously only woke eh thread on resets).
PATCH - Mike Christie [mikenc@us.ibm.com]
             shorten the task management fucntion values
PATCH - Mike Christie [mikenc@us.ibm.com]
             rename the rest of the task enum values by
            prefixing a ISCSI_ to them
PATCH - Mike Christie [mikenc@us.ibm.com]
           This patch now adds scsi-ml template eh functions.
           We do not have any internal recovery so we are at
           the mercy of the scsi-ml eh unjam host strategy
           handler. Becuase of default timeout settings, most of
           the time all that happens is that the session timeout
           occurs and we end up dropping the session. Future
           patches will ensure we always do the session and
           replacement timeout checks, so that for detectable
           transport problems we will not be offlined by the scsi
           error handler.
PATCH - Mike Christie [mikenc@us.ibm.com]
              begin to remove old scsi-eh funtions. These do not
             really end up doing anything anyways.
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm iscsi-recovery.c and iscsi-timer.c. They will be
            replaced with scsi-ml functions in the next patches.
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm disk_command_timeout from driver. This should
            be set from the scsi sysfs attrs.
PATCH - Mike Christie [mikenc@us.ibm.com]
            adds some wrappers around the task refcount
            access, and then moves the completion of the scsi
            commands to when the refcount drops to zero.
            This  has the benefit of not having code outside
             iscsi-task.c peak into the iscsi_task structure to
             try and determine whether
PATCH - Mike Christie [mikenc@us.ibm.com]
             rm find_task_for_cmnd(). It is not used.
1045121 - Amendment to README
               Includes that the sysfsutils package needs to be
               installed for the compilation of the driver.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Remove read queue depth attr (provided by scsi-ml).
PATCH - Mike Christie [mikenc@us.ibm.com]
             Cleanup up some of the nop-in and nop-out code.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Only drop a session if one is established.
PATCH - Mike Christie [mikenc@us.ibm.com]
             Combine iscsi_get_header with iscsi_read_header, and iscsi_get_data
             with iscsi_read_data.
PATCH - Mike Christie [mikenc@us.ibm.com]
             move handle_rx_requests to iscsi-recv-pdu.c
PATCH - Mike Christie [mikenc@us.ibm.com]
             Convert dirver to kthreads, use
            wait_event_interruptible instead of wait_for_session,
            use prepare/finish_wait functions, and replace
            tx_blocked_q with a semaphore
1025807 - Async event support broken with latest code
               Incorporated changes suggested by Mike Christie
 
 
Thanks,
Naveen Burmi.
naveenb@cisco.com