| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
The README said that we auto bind all ifaces in /etc/iscsi/ifaces
but we only autobind ones with the transport name tcp.
|
|
|
|
|
| |
If the session is down we want to force a retry.
DID_TRANSPORT_DISRUPTED is limited to 5, so use DID_IMM_RETRY.
|
|
|
|
|
| |
Add some info using cxgb3i and cxgb3i to the README and
also add some info on the host mode.
|
|
|
|
|
| |
uint64_t can be unsigned long long or unsigned long depending on the
arch. This just has us use sscanf with PRIu64 to read in our uint64_t.
|
|
|
|
| |
Forgot to add files with cxgb3i helpers.
|
|
|
|
|
|
| |
cxgb3i can only support 15360 or smaller. This patch adds a callback
to the transport so that drivers can override the default settings
or do any other setup they may need to do before we try to login.
|
|
|
|
|
|
|
|
|
|
|
| |
We have bind by ip address in io.c, but it was never enabled
because the iface code did not read it in for iscsi_tcp.
Now with offload using the ipaddress field to set the addr of
the port, if the user forgets to the set the transport type
to bnx2i or cxgb3i we will end up in the io.c code trying to
bind to session using bind() which is going to end horribly.
This has us warn the user and fail the connect.
|
|
|
|
|
| |
If we do not recognize the kernel event we should not read in or begin
process to the data payload.
|
|
|
|
|
|
|
| |
Update 2.6.14 - .23 support.
Note: 2.6.16 and below is still broken. Please email the
open-iscsi list if you need this.
|
|
|
|
| |
Update 2.6.24 - .25 compat patch.
|
|
|
|
| |
Update 2.6.26 compat patch.
|
|
|
|
|
|
|
| |
This updates the 2.6.27 kernel patch and it removes
the TARGET_BUSY and starget->can_queue use, because
if bugs in 2.6.27 - 2.6.30 (patches sent
to fix this for 2.6.30 but not yet merged).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This syncs the kernel modules to the upstream commit:
commit 3c0a012b4f02412bc00b4b71b14d273ef579a44e
Author: Mike Christie <michaelc@cs.wisc.edu>
Date: Mon May 11 14:07:49 2009 -0400
libiscsi: add debug printks for iscsi command completion path
This patch just adds some debug statements for the abort
and completion paths.
which was sent for the 2.6.31 feature window.
|
|
|
|
|
|
| |
Fix bad merge. The bound ep connect event number should be 19.
When transferring patches from the offload tree to head this got
set to the wrong value.
|
|
|
|
|
| |
This has idbm_print_discovered use the
idbm_print_node_and_iface_tree and idbm_print_node_flat helpers.
|
|
|
|
|
|
|
|
| |
This creates a default iface for all offload cards. The name will
be "transport_name.hwaddress". The user just has to then set the
ip address by doing
iscsiadm -m iface -I transport_name.hwaddress -o update -n iface.ipaddress -v A.B.C.D
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This prints out the info for iSCSI hosts in the system.
EXAMPLE
-------
iscsiadm -m host
cxgb3i: [9] 10.10.15.50,[00:07:43:05:97:07],eth3 <empty>
$TRANSPORT: [$HOST_NO] $HOST_IP,[$MAC},$NETDEV $INITIATORNAME
EXAMPLE
---------
iscsiadm -m host -P 2
Host Number: 9
State: running
Transport: cxgb3i
Initiatorname: <empty>
IPaddress: 10.10.15.50
HWaddress: 00:07:43:05:97:07
Netdev: eth3
*********
Sessions:
*********
Target: iqn.2001-04.com.home:meanminna
Current Portal: 10.10.15.10:3260,1
Persistent Portal: 10.10.15.10:3260,1
**********
Interface:
**********
Iface Name: cxgb3i
Iface Transport: cxgb3i
Iface Initiatorname: iqn.2005-06.com.redhat:noisymax
Iface IPaddress: 10.10.15.50
Iface HWaddress: 00:07:43:05:97:07
Iface Netdev: eth3
SID: 8
iSCSI Connection State: LOGGED IN
iSCSI Session State: LOGGED_IN
Internal iscsid Session State: NO CHANGE
**********
Interface:
**********
Iface Name: cxgb3i.def
Iface Transport: cxgb3i
Iface Initiatorname: iqn.2005-06.com.redhat:madmax
Iface IPaddress: 10.10.15.50
Iface HWaddress: 00:07:43:05:97:07
Iface Netdev: eth3
SID: 9
iSCSI Connection State: LOGGED IN
iSCSI Session State: LOGGED_IN
Internal iscsid Session State: NO CHANGE
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
iscsiadm was printing the iface's sessions, but to be consistent
with the node mode, we want to print out the records using
the iface.
iscsiadm -m iface -P 1
Iface: default
Target: iqn.2001-04.com.home:meanminna
Portal: 20.15.0.9:3260,1
Portal: 10.10.15.10:3260,1
Iface: iser
Iface: cxgb3i.def
Iface: cxgb3i
Target: iqn.2001-04.com.home:meanminna
Portal: 10.10.15.10:3260,1
iscsiadm -m node -P 1
Target: iqn.2001-04.com.home:meanminna
Portal: 20.15.0.9:3260,1
Iface Name: default
Portal: 10.10.15.10:3260,1
Iface Name: cxgb3i
Iface Name: default
To get the list of sessions running through a host, we want
to add a host mode. This would work like the session mode, but
at the host level.
|
|
|
|
|
| |
If a value is not set we are returning either defaut or unknown
or <empty>. This unifies us to use <empty>.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If we are using a offload card, then iface_set_param will match the
iface info to a scsi_host and pass that info down to setup the net
settings of the port (currently we just set the ip address). When we
create the tcp/ip connection by calling ep_connect, we currently just go
by the routing table info.
I think there are two problems with this.
1. Some drivers do not have access to a routing table. Some drivers like
qla4xxx do not even know about other ports.
2. If you have two initiator ports on the same subnet, the user may have
set things up so that session1 was supposed to be run through port1. and
session2 was supposed to be run through port2. It looks like we could
end with both sessions going through one of the ports.
Also how do you edit the routing table for the offload cards? You cannot
use normal net tools like route can you?
3. If we set up hostA in the iface_set_param step, but then the routing
info leads us to hostB, we are stuck.
I did the attached patches to fix this. Basically we just pass down the
scsi host we want to go through.
|
|
|
|
| |
sysfs_file is never used.
|
|
|
|
| |
Adding missing forward declaration.
|
|
|
|
|
| |
iscsidm -m fw was printing out the %d for the conn
params instead of the cid.
|
|
|
|
|
|
| |
If we were to find more thatn IBFT_MAX targets we would
overrun the tgt or nic arrary. This just has us spit an
error and return.
|
|
|
|
|
|
| |
If the ibft functions are called multiple times tgt_cnt and
nic_cnt will continue to increment, so they need to be cleared
each time.
|
|
|
|
|
|
|
| |
From Hans De Goede:
ip-addr will not exist when using dhcp, so raising an error
when it does not exist is wrong.
|
|
|
|
|
|
|
| |
Multiple fixes to ppc boot code from Hans de Goede:
- nic and dev count need to be cleared. When called multiple times from
the lib, they would overrun the nic/initiator arrays.
- Have fwparam_ppc not set global dev_count when calling find_file.
|
|
|
|
|
|
|
|
|
|
|
|
| |
If you were to run
iscsiadm -m discovery -ip -l
to try and login to a targets found during discovery using the
/etc/iscsi/send_targets/ip settings, and if CHAP was used
then the discovery session login would fail. The reason for
this is because we were using the password_len fields to specify
the password length to try and allow for any chars.
This patch adds those idbm fields back.
|
|
|
|
| |
This sets node.session.queue_depth for the session's devices.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- fix double close in fwparam ppc
- fix ofdevs and devtree leaks
- fix fwparam filename init
- fix fwparam ibgt null free
- fix iscsistart fw boot and info (sysfs init was not yet called)
- fix iscsistart startup
iscsid login thread may not be started when iscsistart sends
login request. Patch 5b9b71007e43e0c930872b3cadd5c7434a54539d
had removed the wait when ECONNREFUSED is returned, which
added the regression. TODO: Should add a wait/wakeup for this
instead of simple loop and sleep.
|
|
|
|
|
|
|
|
|
|
|
| |
From Dotan Barak:
Add missing ';' in backport patches.
When enabling debug in compilation, the lack of semicolon caused a
compilation
failures on older kernels.
Signed-off-by: Dotan Barak <dotanba@gmail.com>
|
|
|
|
|
|
|
|
| |
Increase default node.conn[0].iscsi.MaxRecvDataSegmentLength
to 262144 to closer match IO sizes we see from block layer.
The default block layer max is actually 512K, but in testing
setting the MaxRecvDataSegmentLength to this high a value
caused perf regressions with some targets.
|
|
|
|
|
|
|
| |
This sets the default xmit thread nice to -20. This will
increase throughput and decrease latency, but could hog
the cpu more. In the latter case you can set the nice in
/etc/iscsi/iscsid.conf's node.session.xmit_thread_priority setting.
|
|
|
|
| |
Run spell checker over README.
|
|
|
|
| |
Fix up spelling of interface in README found by Ulrich.
|
|
|
|
|
|
| |
mv sysfs cache cleanup so it runs after the actor test, because
software iscsi could run for a while without ever firing a kernel
or mgmt event, but the sysfs layout could still change.
|
|
|
|
|
| |
The wiki is down and probably dead for good. This removes a bad
link to it.
|
|
|
|
|
| |
There is a new module, libiscsi_tcp, that is installed. This
updates the README.
|
|
|
|
|
|
|
| |
The event code does not belong in the mgmt ipc code, because
the event code now handles mgmt ipc, requests, actor events,
isns events, kernel ipc events. This moves the code to a new
file.
|
|
|
|
|
|
| |
After handling an event we should flush the sysfs cache because
the device links could change or the attr values could change due
to handling the event.
|
|
|
|
| |
Update version to 871.
|
|
|
|
|
|
| |
When the offload iface settings changed the doc page got
misedited. This should fix it and add some info on what
the hw address is.
|
|
|
|
|
| |
The connection failed message did not say which connection
failed. This has it print out the portal and a strerror value.
|
|
|
|
| |
When doing a make notify user that iscsistart was built.
|
|
|
|
| |
This has us use a safer strlcpy instead of strncpy when we can.
|
|
|
|
|
| |
We messed up strncat in so many different ways that strlat could
fix. This patch converts the userspace uses of strncat to strlcat.
|
|
|
|
|
|
|
|
| |
Separate out the build of "kernel:" and "user:" targets.
[Mike please revisit the @echo output if we need anything added]
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
|
|
|
|
| |
Fix compat patch that was broke from the eh timed out patch.
|
|
|
|
|
|
|
| |
Merge upstream fix for command timeout oops
(iscsi command does not oops, but if a command from another
driver timesout that will oops, because iscsi was missetting
the default transportt to its time out handler).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Patch and mail from Hans De Goede:
Hi,
While testing I noticed that "iscsiadmin -m fw" does not work properly
on newer
(rawhide atleast) kernels, the attached patch (already applied to the
Fedora
devel packages) fixes this.
Regards,
Hans
|