- Changed return type of srpt_release_channel() from int to void
because the return value did depend in a trivial way on the function
arguments. Changed the return type of srpt_find_and_release_channel()
and several of its callers also from int to void. Instead
srpt_cm_handler() now directly sets its own return value.
- Converted a list_for_each_entry_safe() into a list_for_each_entry() loop
because no elements are added or deleted in the loop body.
- Converted debug-printk() statements into TRACE_DBG().
- Changed indentation of one statement from 12 to 16 (was not reported by
checkpatch).
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@1003 d57e44dd-8a1f-0410-8b47-8ef2f437770f
written into a buffer with a size of 32 bytes. Enlarged the buffer
to 36 bytes.
- Converted sprintf() calls to snprintf() calls.
- Added more TRACE_ENTRY() / TRACE_EXIT() statements.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@992 d57e44dd-8a1f-0410-8b47-8ef2f437770f
/proc/scsi_tgt/ib_srpt/trace_level such that the behavior of the
ib_srpt target is consistent with that of other SCST drivers.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@991 d57e44dd-8a1f-0410-8b47-8ef2f437770f
wait until srpt_refresh_port_work() finished. This race condition could be
triggered during module removal.
- Added a kernel module parameter called "trace_flag" that allows to set the
trace flags for the ib_srpt module before module initialization starts.
- Added sysfs variable /sys/class/infiniband_srpt/trace_level that allows to
display and to modify the enabled trace flags in a human-readable form.
- Added several TRACE_ENTRY() / TRACE_EXIT() statements.
- Added more comments.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@988 d57e44dd-8a1f-0410-8b47-8ef2f437770f
- RHEL 5 + OFED: fixed kernel oops that was triggered by the reception of
an asynchronous InfiniBand event.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@880 d57e44dd-8a1f-0410-8b47-8ef2f437770f
call to scst_unregister() in srpt_remove_one(). Result:
- The kernel oops triggered by loading the ib_srpt kernel module before
the InfiniBand subnet manager became active is now fixed.
- It is now possible to log in via the SRP protocol through HCA's that
became active after the ib_srpt kernel module finished module
initialization.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@878 d57e44dd-8a1f-0410-8b47-8ef2f437770f
handler registration: sport->mad_agent is now reset to NULL when MAD
handler registration fails.
- Fixed bug in srpt_refresh_port(): this function now returns a negative
value instead of zero when MAD registration fails.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@877 d57e44dd-8a1f-0410-8b47-8ef2f437770f
- Simplified this Makefile: the target all now directly refers to
src/Module.symvers. Removed the phony target Module.symvers.
- An error message is now printed if OFED has been built with srpt=y.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@817 d57e44dd-8a1f-0410-8b47-8ef2f437770f
when an SRP login request is rejected. This fixes a memory leak: the
InfiniBand core (drivers/infiniband/core/cm.c) only calls cm_free_work()
when the registered cm_handler function, i.e. srpt_cm_req_recv()
returns an error code. This patch has been submitted by by Frank Zago.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@800 d57e44dd-8a1f-0410-8b47-8ef2f437770f
ports. show_login_info() produces the information that can be retrieved
from /sys/class/infiniband_srpt/srpt-*/login_info. Note: as of today
there do not yet exist any InfiniBand HCA's with more than two ports.
Thanks to Frank Zago for reporting this issue.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@799 d57e44dd-8a1f-0410-8b47-8ef2f437770f
I replaced all occurences of DISTDIR with DESTDIR and also introduced SBINDIR (and friends) to some other Makefiles.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@719 d57e44dd-8a1f-0410-8b47-8ef2f437770f
