- __attribute__((aligned(sizeof(long))) added to all field with different protection than the previous field to make sure they don't share the same bus line. Just in case, actually, compiler should do the alignment by default.
- Small docs updates.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@1948 d57e44dd-8a1f-0410-8b47-8ef2f437770f
there seems to be a problem with the Linux SCSI stack because when we delete a
LUN on the target, it does not get deleted from Linux. Still investigating.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@1947 d57e44dd-8a1f-0410-8b47-8ef2f437770f
TRACE_DBG("Created tid '%08lX'", (unsigned long)&tr_id[4]);
is wrong. All it will do is print the address of the fifth byte in the tr_id.
It should cast to an unsigned long * and deref it, at least. However, I have
anothe change that prints out the tr_id as a SAS address.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@1934 d57e44dd-8a1f-0410-8b47-8ef2f437770f
use an OUI that I think is not currently in use, so it seems unlikely that we
will clash with anyone else, but you never know. I might have to change it
later.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@1928 d57e44dd-8a1f-0410-8b47-8ef2f437770f
The individual patches are a subset of the earlier ones, with
only line number changes to apply cleanly.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@1922 d57e44dd-8a1f-0410-8b47-8ef2f437770f
- Made sure that in mode thread=1 not only send completions but also receive completions are processed in thread context.
- Merged IB send and receive completion queues again such that both types of completions are processed in a fairly.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@1904 d57e44dd-8a1f-0410-8b47-8ef2f437770f
- Changed default value for the kernel module parameter thread from 0 to 1 because mode 0 can trigger soft lockup complaints under sufficient load.
- Changed default value for the kernel module parameter srpt_autodetect_cred_req from false to true.
Non-functional changes:
- Removed several superfluous casts in ib_srpt.c.
Performance-related changes:
- Rearranged if/then/else statements such that the most frequently executed code occurs first.
- When SCST_MAX_TGT_DEV_COMMANDS > 128, the SRP receive queue size is now SCST_MAX_TGT_DEV_COMMANDS instead of 128.
Bug fixes:
- Fixed performance bug in modes thread=0 and 2.
- Fixed a bug in IB send queue overflow handling.
- Fixed race related to IB completion thread creation (mode thread=1).
- Made sure that the following warning is no longer triggered during module removal:
WARNING: at fs/proc/generic.c:825 remove_proc_entry+0x261/0x2f0()
Call Trace:
[<ffffffff8104d74b>] warn_slowpath_common+0x7b/0xc0
[<ffffffff8104d7f1>] warn_slowpath_fmt+0x41/0x50
[<ffffffff8116e8a1>] remove_proc_entry+0x261/0x2f0
[<ffffffff810832c8>] ? debug_check_no_locks_freed+0xc8/0x150
[<ffffffff81082ecb>] ? mark_held_locks+0x6b/0xa0
[<ffffffff8110f95f>] ? kfree+0xcf/0x180
[<ffffffff810831a5>] ? trace_hardirqs_on_caller+0x145/0x190
[<ffffffff810831fd>] ? trace_hardirqs_on+0xd/0x10
[<ffffffffa05835c2>] srpt_unregister_procfs_entry+0x22/0x40 [ib_srpt]
[<ffffffffa0583aba>] srpt_release+0x8a/0x1a0 [ib_srpt]
[<ffffffffa04cbcc0>] scst_unregister_target+0x70/0x3f0 [scst]
[<ffffffff811bf6d7>] ? kobject_put+0x27/0x60
[<ffffffff8126ff67>] ? put_device+0x17/0x20
[<ffffffffa058587e>] srpt_remove_one+0xee/0x1a0 [ib_srpt]
[<ffffffffa03ddfc5>] ib_unregister_client+0x65/0x120 [ib_core]
[<ffffffffa0588f43>] srpt_cleanup_module+0x5f/0xc3 [ib_srpt]
[<ffffffff8108e8dd>] sys_delete_module+0x1bd/0x280
[<ffffffff8135e5ad>] ? retint_swapgs+0xe/0x13
[<ffffffff810831a5>] ? trace_hardirqs_on_caller+0x145/0x190
[<ffffffff810030db>] system_call_fastpath+0x16/0x1b
---[ end trace 3fc26663490f6a89 ]---
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@1901 d57e44dd-8a1f-0410-8b47-8ef2f437770f
