Speeds up reading from a RAM disk via ib_srpt by about 30%.
Tested with iSCSI-SCST, ib_srpt and scst_local.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
In it:
- Fixed NULLIO, which the original patch breaks
- Changed on all fast paths sBUG_ON() on EXTRACHECKS_BUG_ON()
- Fixed errors path in non_fileio_exec() and fileio_alloc_data_buf().
- Renamed zero_copy_read to just zero_copy.
- Docs and change log updated
- Some cleanups
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@4213 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Fix three occurrences of the following warnings reported by the checkpatch
script included with Linux kernel version 3.3:
WARNING: min() should probably be min_t()
WARNING: max() should probably be max_t()
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@4172 d57e44dd-8a1f-0410-8b47-8ef2f437770f
This patch:
* fixes signatures for log_debug and log_pdu macros
* collapses __log_<LEVEL> functions into one
* adds conditional logging
Here conditional logging means that the logging priority and logging level may
vary depending on some dynamic parameters. This is helpful when software is
able to recover from errors which occur periodically and hence, there is no
need to pollute system logs with tons of repetitive non-critical lines.
The following macros are added to support this feature:
* log_info_cond
* log_warning_cond
* log_error_cond
They take an additional first parameter called "level". If it is 0 then they
act as their non-cond counterparts, otherwise they act like log_debug with the
specified level.
Those who are not interested in conditional logging are not affected: they can
continue using plain old logging macros without any changes.
Signed-off-by: Sergey Myasnikov <tigra564@gmail.com>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@4161 d57e44dd-8a1f-0410-8b47-8ef2f437770f
This patch moves the following utility functions from iscsi_scstd.c to misc.c
and makes them public (i.e. non-static):
* set_non_blocking
* sock_set_keepalive
Signed-off-by: Sergey Myasnikov <tigra564@gmail.com>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@4160 d57e44dd-8a1f-0410-8b47-8ef2f437770f
__attribute__((aligned(x))) into __aligned(x) in kernel code since
checkpatch complains about the former.
BSD-signed-off-by: Bart Van Assche <bvanassche@acm.org>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@3952 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Issue:
------------
The DevRegAttribute is sending a different Source Attribute (ISCSI name) which
represents the Storage Node for each target. So Microsoft isns server considers
the last request as the storage node and doesn't discover the previous targets.
Root Cause
----------------
Function ISNS_FUNC_DEV_ATTR_REG
Target A
Source attribute
ISCSI-Name = <IQN A Name>
Operating Attribute
ISCSI-Name = <IQN A Name>
Flags
Replace Access = 1
Target B
Source attribute
ISCSI-Name = <IQN B Name>
Operating Attribute
ISCSI-Name = <IQN B Name>
Replace Access = 1
So as per the request format Microsoft isns Server always takes the last target
as the source node. and register the same. Hence it doesn't discover the
targets other than last one. The Source Attribute has to be fixed for all targets. It
would be better if we use the first target iscsi name (rather than the last target
iscsi name) as source attribute iscsi name to represent the storage node.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@3863 d57e44dd-8a1f-0410-8b47-8ef2f437770f
from the previous state. Hence, the corresponding labels were renamed to
2debug, 2release and 2perf.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@3736 d57e44dd-8a1f-0410-8b47-8ef2f437770f
into a new macro called wait_event_locked().
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
+
Micro-optimize wait_event_locked(). Switch from default_wake_function() to
autoremove_wake_function() (via DEFINE_WAIT()). The result is that
wait_queue_head_t.lock is locked once instead of twice for wakeup + wait queue
removal.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@3591 d57e44dd-8a1f-0410-8b47-8ef2f437770f
scst, iscsi-scst and qla2x00t. This makes it unnecessary hard for users to
run several target drivers simultaneously. Additionally, several variants of
these init scripts exist for different distributions. This patch unifies all
these init scripts into a single init script. Additionally, some new features
have been added:
- Kernel module parameters can now be specified per kernel module in
/etc/default/scst. An example:
ib_srpt_parameters="thread=1 srp_max_req_size=4200"
- When using the sysfs interface, compute the kernel module list from scst.conf.
Or, editing the SCST_MODULES variable is no longer necessary.
- When using the procfs interface, the SCST handler kernel module list is
computed from /etc/scst.conf. The list of target driver kernel modules must
now be configured in /etc/default/scst.
Other changes:
- scstadmin output is shown if scstadmin fails.
- Running make -C scstadmin install does no longer cause the scst script to
be started at the next reboot - it is better to leave this as an explicit step.
- Made the error message generated if the lsb-core package has not yet been
installed more clear.
- Use proper init script path on Slackware.
- make uninstall: run remove_initd under chroot if $(DESTDIR) is not empty.
- Fixed a bug in "make -C scstadmin uninstall": the path specified for removing
/etc/init.d/scst was wrong.
- The command /etc/init.d/scst restart will only try to start SCST if stopping
succeeded.
- Replaced "mkdir -m 755 -p $(DESTDIR)$(INITDIR)" by
"install -d $(DESTDIR)$(INITDIR)" because of uniformity.
These changes have been tested on Ubuntu 11.04, Fedora 15, CentOS 5.6,
Scientific Linux 6.0, openSUSE 11.4, SLES11 SP1 and Slackware 13.37.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@3564 d57e44dd-8a1f-0410-8b47-8ef2f437770f
