Whamcloud - gitweb
fs/lustre-release.git
19 years ago* tidied up NID printing (s/LPX64/%s/ && s/nid/libcfs_nid2str(nid)/)
eeb [Mon, 12 Sep 2005 13:49:36 +0000 (13:49 +0000)]
*   tidied up NID printing (s/LPX64/%s/ && s/nid/libcfs_nid2str(nid)/)

19 years ago* Cleaned up portals compatibility tests into a couple of inlines
eeb [Sun, 11 Sep 2005 13:54:35 +0000 (13:54 +0000)]
*    Cleaned up portals compatibility tests into a couple of inlines
     in lib-lnet.h

*    Added (but didn't test) portals compatibility support for
     gm, openib and ra.

19 years ago* Added check for portals compatibility mode in LNDs that don't support it
eeb [Sat, 10 Sep 2005 17:12:16 +0000 (17:12 +0000)]
*   Added check for portals compatibility mode in LNDs that don't support it
    yet.

19 years ago* Got vibnal LNET/portals wire compatibility working
eeb [Sat, 10 Sep 2005 17:05:04 +0000 (17:05 +0000)]
*   Got vibnal LNET/portals wire compatibility working

*   Removed bad LPSZ in format strings (I got rid of lnet_size_t)

*   Changed vibnal NID printing from LPX64 to %s(libcfs_nid2str(nid))

19 years ago* LNET/portals wire compatibility working on elan and tcp. Set the lnet
eeb [Sat, 10 Sep 2005 03:48:48 +0000 (03:48 +0000)]
*   LNET/portals wire compatibility working on elan and tcp.  Set the lnet
    module parameter "portals_compatibility" to...

    "strong"    Compatible with portals and LNET "strong" and "weak"

    "weak"      Compatible with any value of LNET portals_compatibility

    "none"      Compatible with LNET "weak" and "none". This is the default.

    Old XML and existing old configuration profiles (logs) can be used as-is.

*   Updated GM README

*   Backed out most of the change to lconf that used hostaddr to construct the
    LNET NID.  It now signals an error if the XML contains > 1 --hostaddr, or
    if the --hostaddr doesn't match the NID, since it's likely manual
    intervention will be required in these cases.

19 years agooptional build for portals.
liangzhen [Fri, 9 Sep 2005 15:40:08 +0000 (15:40 +0000)]
optional build for portals.
1. Portals will not be built by default
2. To build portals: configure --with-portals=yes .....

19 years ago* Fix ogdb-host file generation, pickup the correct lnet modules as well as the legac...
pjkirner [Thu, 8 Sep 2005 20:37:10 +0000 (20:37 +0000)]
* Fix ogdb-host file generation, pickup the correct lnet modules as well as the legacy portals modules (for testing purposes only)

19 years ago* Added GM README
eeb [Thu, 8 Sep 2005 17:18:24 +0000 (17:18 +0000)]
*   Added GM README

19 years ago* Removed unused parameters from LNet??? APIs (e.g. interface handle)
eeb [Thu, 8 Sep 2005 15:18:55 +0000 (15:18 +0000)]
*  Removed unused parameters from LNet??? APIs (e.g. interface handle)

*  Removed unused LNet??? APIs.

*  Removed many scalar typedefs inherited from portals.

*  fixed up alignment in some decls that s/ptl/lnet/ had unaligned.

*  updated sanity.sh to s/portals.debug/lnet.debug/

*  verified lnet can zeroconf mount a pre-lnet filesystem after
   lctl --write_config <pre-lnet-xml>

19 years agoSmallfix for function define.
liangzhen [Wed, 7 Sep 2005 13:33:08 +0000 (13:33 +0000)]
Smallfix for function define.

19 years ago* Added support for routable RDMA-ed REPLY messages to qswlnd
eeb [Tue, 6 Sep 2005 09:30:54 +0000 (09:30 +0000)]
*    Added support for routable RDMA-ed REPLY messages to qswlnd

19 years ago* ptllnd: added .cvsignore
eeb [Tue, 6 Sep 2005 07:42:56 +0000 (07:42 +0000)]
*   ptllnd: added .cvsignore

19 years ago* Removed nal_{send,recv}_pages() LND APIs (send and receive are passed
eeb [Mon, 5 Sep 2005 19:21:16 +0000 (19:21 +0000)]
*   Removed nal_{send,recv}_pages() LND APIs (send and receive are passed
    either VM frags (iov != NULL) or page frags (kiov != NULL) but not both.

*   Ensure that the order of networks declared in the "networks" and "routes"
    breaks ties when determining which peer NID to use.

19 years ago* Added hopcounts to route table
eeb [Fri, 2 Sep 2005 18:56:56 +0000 (18:56 +0000)]
*   Added hopcounts to route table

*   Changed ptlrpc_uuid_to_peer() to choose the matching UUID with the shortest
    hopcount

*   Changed lconf to use a single UUID string for all target NIDs, so the
    client can choose which one to use at runtime.

*   Stripped out all the unused network configuration stuff from lconf

19 years agoAdd PTLLND to the LND enum.
pjkirner [Fri, 2 Sep 2005 12:52:24 +0000 (12:52 +0000)]
Add PTLLND to the LND enum.

19 years agoPatch for Lustre Networking Reorganization
liangzhen [Fri, 2 Sep 2005 09:32:11 +0000 (09:32 +0000)]
Patch for Lustre Networking Reorganization

DONE:
1. Fixing of building both lnet and portals
2. Fixing of conflicting symbols in lnet and portals
   - exported APIs of lnet/libcfs with name like ptl_* are renamed to libcfs_*
   - exported APIs of lnet/lnet with name like ptl_* are renamed to lnet_*
   - exported APIs of portals/libcfs with name like libcfs_* are renamed to libptl_*
   - modules name of portals/libcfs/libcfs.ko to portals/libcfs/libptl.ko
3. /proc entry for lnet is /proc/sys/lnet
4. Listen port of socklnd is 988, listen port of socknal is 989
5. Pseudo device for lnet is /dev/lnet
6. Fixing of build lnet/klnds/ptllnd
7. Fixing of module path and /proc path in lnet/utils lustre/utils lustre/tests

TODO:
1. Renaming of unexported symbols in lnet.
2. Renaming of types and macro
3. Add option for building portals
4. Misc fix and testing

19 years agoAdded build infrastructre for PTLLND.
pjkirner [Thu, 1 Sep 2005 16:21:57 +0000 (16:21 +0000)]
Added build infrastructre for PTLLND.
Plus dummy PTLLND that basic interactions between LNET and PORTALS.

19 years agoReorganize LNET API files, so that PTLLND
pjkirner [Thu, 1 Sep 2005 15:16:49 +0000 (15:16 +0000)]
Reorganize LNET API files, so that PTLLND
can include both LNET and PORTALS.

19 years agoFix missing NAL->LND
pjkirner [Thu, 1 Sep 2005 13:02:53 +0000 (13:02 +0000)]
Fix missing NAL->LND

19 years agoSmallfix for lnet build
liangzhen [Thu, 1 Sep 2005 04:11:07 +0000 (04:11 +0000)]
Smallfix for lnet build

19 years agoChanges necessary to make liblustre build after LNET rename
pjkirner [Thu, 1 Sep 2005 03:46:00 +0000 (03:46 +0000)]
Changes necessary to make liblustre build after LNET rename

19 years agoChanges for LNET rename of NAL -> LND
pjkirner [Thu, 1 Sep 2005 02:52:35 +0000 (02:52 +0000)]
Changes for LNET rename of NAL -> LND

19 years agoRename Directories in LNET
pjkirner [Thu, 1 Sep 2005 00:21:57 +0000 (00:21 +0000)]
Rename Directories in LNET
knals -> klnds
unals -> ulnds
And associated build fixes.

19 years ago * Applied Andreas' tcpnal compiler optimization bugfix patch
eeb [Wed, 31 Aug 2005 21:34:02 +0000 (21:34 +0000)]
  *   Applied Andreas' tcpnal compiler optimization bugfix patch
      from HEAD portals (different way of constructing tcp HELLO
      header to avoid pointer aliasing) to lnet

  *   Applied qswnal build fix to lustre-portals.m4 from HEAD
      portals to lnet

  *   lnet version of gmnal running @ HP

  *   fixed bad 64bit cast in acceptor.c

  *   fixed lconf to work with newconfig modules under lnet

19 years ago* Applied implicit_loopback fixes to portals (it was previously
eeb [Wed, 31 Aug 2005 12:51:50 +0000 (12:51 +0000)]
*   Applied implicit_loopback fixes to portals (it was previously
    applied to lnet)

*   Minor formatting changes to lnet/lnet/{lib-move,router}.c

19 years agoFixes for LNET Directory rename
pjkirner [Wed, 31 Aug 2005 03:32:53 +0000 (03:32 +0000)]
Fixes for LNET Directory rename

19 years ago* fixed implicit_loopback bug
eeb [Tue, 30 Aug 2005 20:36:52 +0000 (20:36 +0000)]
*    fixed implicit_loopback bug

19 years agoLNET Reorg Phase 1
pjkirner [Tue, 30 Aug 2005 20:18:33 +0000 (20:18 +0000)]
LNET Reorg Phase 1
r=pjk
r=eeb

19 years agoPhase 1 of LNET Reorg
pjkirner [Tue, 30 Aug 2005 20:15:51 +0000 (20:15 +0000)]
Phase 1 of LNET Reorg
r=pjk
r=eeb

19 years agoPortals -> LNet:
liangzhen [Wed, 17 Aug 2005 14:16:06 +0000 (14:16 +0000)]
Portals -> LNet:
rename Ptl_* APIs to LNet_*

19 years agomgs operations
lincent [Mon, 15 Aug 2005 18:24:27 +0000 (18:24 +0000)]
mgs operations

19 years ago* iibnal and openibnal packed wire format updates from HEAD portals
eeb [Wed, 10 Aug 2005 11:04:18 +0000 (11:04 +0000)]
*  iibnal and openibnal packed wire format updates from HEAD portals

19 years ago* 7164 fix newconfig acceptor startup assertion
eeb [Sun, 31 Jul 2005 13:32:25 +0000 (13:32 +0000)]
*    7164 fix newconfig acceptor startup assertion

19 years agob=6663
nathan [Mon, 25 Jul 2005 23:55:57 +0000 (23:55 +0000)]
b=6663
adding lvfs_mount, begin changes to llite_lib.c

19 years agoBranch b_hd_newconfig
nathan [Thu, 14 Jul 2005 23:29:44 +0000 (23:29 +0000)]
Branch b_hd_newconfig
Remove the "-O2" requirement for htonl references

19 years ago* better ranal error message on FMA CQ overflow
eeb [Mon, 11 Jul 2005 16:43:47 +0000 (16:43 +0000)]
*   better ranal error message on FMA CQ overflow

*   fixed socknal modparam that should have been RO under /proc or /sys

*   skip the loopback interface in ptl_set_ip_niaddr() when no interfaces have
    been specified.

*   #if 0 out the 'allow_destination_aliases' portals module param.  It's too
    hard to describe to punters.

*   change the meaning of the 'implicit_loopback' portals module parameter to
    be more precisely just that.

*   don't fail when adding a duplicate route; just ignore it.

19 years ago* More intuitive newconfig acceptor module params
eeb [Fri, 8 Jul 2005 16:39:24 +0000 (16:39 +0000)]
*   More intuitive newconfig acceptor module params

19 years ago* newconfig liblustre/tcpnal runs sanity OK
eeb [Fri, 8 Jul 2005 14:51:27 +0000 (14:51 +0000)]
*    newconfig liblustre/tcpnal runs sanity OK

19 years ago* first signs of life from newconfig tcpnal + liblustre echo_client
eeb [Thu, 7 Jul 2005 17:51:53 +0000 (17:51 +0000)]
*   first signs of life from newconfig tcpnal + liblustre echo_client

19 years ago* socknal working again
eeb [Thu, 7 Jul 2005 10:04:03 +0000 (10:04 +0000)]
*   socknal working again

19 years ago* socknal: removed unused var
eeb [Wed, 6 Jul 2005 15:54:56 +0000 (15:54 +0000)]
*   socknal: removed unused var

19 years ago* Compiles after merging b1_4
eeb [Tue, 5 Jul 2005 18:35:31 +0000 (18:35 +0000)]
*   Compiles after merging b1_4

*   Changed socknal...
    - use PID as well as NID to match connections so userspace (tcpnal) clients
      can be distinguished without changing the NID format.
    - unprivileged port == userspace client
    - don't create new connections to userspace clients
    - derive the NID/PID of a userspace client from the remote IP/port

*   Changed tcpnal...
    - use non-privileged ports
    - no concept of own NID (peer assigns)
    - don't accept connections

19 years ago* Changed default vib tx descriptor pool sizes
eeb [Fri, 1 Jul 2005 09:33:29 +0000 (09:33 +0000)]
*   Changed default vib tx descriptor pool sizes

19 years ago* More work on config error messages + some code restructure to trap config
eeb [Thu, 30 Jun 2005 17:33:32 +0000 (17:33 +0000)]
*  More work on config error messages + some code restructure to trap config
   errors better.

*  Fixed bug in 'implicit loopback'

*  Completely removed duplicate interface usage checks from generic network
   setup.  The NAL should be doing this.   Currently socknal is the only one
   that supports multiple instances and it doesn't mind at all if 2 different
   network instances use the same interfaces, and that's dead useful for
   testing.

*  Tightened up lonal refcounting; there was a remote chance of a shutdown
   race.

19 years ago* Improved router error messages and move some of the checking out of qswnal
eeb [Wed, 29 Jun 2005 21:33:57 +0000 (21:33 +0000)]
*   Improved router error messages and move some of the checking out of qswnal
    and socknal and into the router itself.

*   Don't force FMR unmap on error if it wasn't mapped in the first place.

*   Make socknal close the incoming connection if it receives junk for
    forwarding

*   allow "lctl network down" == "lctl network unconfigure" (I kept wanting to
    type 'down' and now I can :)

19 years ago* started getting tcpnal working with portals newconfig
eeb [Tue, 28 Jun 2005 16:13:16 +0000 (16:13 +0000)]
*   started getting tcpnal working with portals newconfig

19 years ago* vibnal: ensure FMR is flushed on error
eeb [Tue, 28 Jun 2005 09:43:53 +0000 (09:43 +0000)]
*   vibnal: ensure FMR is flushed on error

19 years agoAdded router.h to EXTRA_DIST to fix 'make rpms'
behlendo [Mon, 27 Jun 2005 23:47:52 +0000 (23:47 +0000)]
Added router.h to EXTRA_DIST to fix 'make rpms'

20 years ago* Added console error to report acceptor startup failure
eeb [Tue, 21 Jun 2005 11:56:32 +0000 (11:56 +0000)]
*   Added console error to report acceptor startup failure

20 years ago* debugged single acceptor
eeb [Tue, 21 Jun 2005 11:48:33 +0000 (11:48 +0000)]
*   debugged single acceptor

*   fixed PtlGetId() to skip the loopback NI

20 years ago* first cut single acceptor for SOCKNAL, RANAL, OPENIBNAL, with
eeb [Mon, 20 Jun 2005 20:23:42 +0000 (20:23 +0000)]
*    first cut single acceptor for SOCKNAL, RANAL, OPENIBNAL, with
     backwards compatibility.

20 years ago* Updated vibnal from b1_4 to use FMR
eeb [Fri, 17 Jun 2005 09:29:28 +0000 (09:29 +0000)]
*  Updated vibnal from b1_4 to use FMR

20 years ago* vibnal: added module parameters for some low-level IB tunables (bug #6443)
eeb [Sat, 11 Jun 2005 13:15:26 +0000 (13:15 +0000)]
*   vibnal: added module parameters for some low-level IB tunables (bug #6443)

*   vibnal: changed some default tunables (bug #6474)

20 years ago* noticed difference between userspace and kernel x86_64 __u64 etc typdefs
eeb [Wed, 8 Jun 2005 19:04:22 +0000 (19:04 +0000)]
*   noticed difference between userspace and kernel x86_64 __u64 etc typdefs
    and adjusted LPU64 etc in <libcfs/linux/kp30.h> appropriately

*   libcfs_pause() -> inline cfs_pause() in <libcfs/linux/linux-prim.h> and
    made more use of it

*   cleaned up NAL loading/instantiated console messages

*   fixed reconnect_interval in connection NALs, where I'd left out * HZ

*   made ???_npeers atomic again; using ???_global lock round the decrement in
    ???_destroy_peer() could deadlock.

*   fixed uninitialised list in vib:kibnal_peer_connect_failed()

*   changed to "nicer" format for /proc/sys/portals/routes

*   more x86_64 testing

20 years ago* Got rid of portals/router; getting it to build in portals/portals using
eeb [Tue, 7 Jun 2005 23:18:56 +0000 (23:18 +0000)]
*  Got rid of portals/router; getting it to build in portals/portals using
   symlinks didn't work on 2.6 and router.c has changed enough for the diffs
   not to make sense any more (and life is too short).

*  reworked how reconnect timeouts work in all connection based NALs, to close
   out a bug where reconnection could be delayed until jiffies wraps.

*  tested vibnal with a single echo client + server @ LLNL on 2.6, x86_64

*  make configure-on-demand rather than configure-on-load the default; modprobe
   in 2.6 does locking and deadlocks if the module it is loading tries to load
   more modules in its init sequence.

*  removed some unused vars etc caught by better warnings when building for
   2.6/x86_64

20 years ago* brought openibnal into the newconfig fold
eeb [Mon, 6 Jun 2005 21:22:25 +0000 (21:22 +0000)]
*  brought openibnal into the newconfig fold

*  moved socket address query function out of NALs and into libcfs

20 years ago* merged 6361 vibnal fixes (IBNAL_OUS_DST_RD = 1, HCA names)
eeb [Fri, 3 Jun 2005 11:20:52 +0000 (11:20 +0000)]
*    merged 6361 vibnal fixes (IBNAL_OUS_DST_RD = 1, HCA names)

*    vibnal to use HCA name as the interface name
        - add hca and IPoIB interface name prefixes to tunables
        - use 1:1 HCA number and IPoIB interface number to determine IP address
          of given HCA

20 years ago* merged in 5858 fix (vibnal arp retries)
eeb [Fri, 3 Jun 2005 09:03:44 +0000 (09:03 +0000)]
*   merged in 5858 fix (vibnal arp retries)

20 years ago* vibnal takes a single interface name which defaults to 'ipoib0'
eeb [Thu, 2 Jun 2005 17:57:59 +0000 (17:57 +0000)]
*    vibnal takes a single interface name which defaults to 'ipoib0'

20 years ago* moved #defines in qswnal.h around to separate fixed constants
eeb [Thu, 2 Jun 2005 07:09:25 +0000 (07:09 +0000)]
*    moved #defines in qswnal.h around to separate fixed constants
     from ones that can be set via modparams

*    placeholder vibnal for newconfig; still need to get the IPoIB
     IP address somehow.

20 years ago* Added interface and socket queries to libcfs (libcfs/linux-tcpip.c). Some
eeb [Wed, 1 Jun 2005 21:23:21 +0000 (21:23 +0000)]
*    Added interface and socket queries to libcfs (libcfs/linux-tcpip.c).  Some
     of this came from socknal_lib-linux.c but was generalised a little so all
     the NALs that use TCP/IP to do connection establishment can use it.

     CAVEAT EMPTOR! Just the linux versions are done; the darwin versions are
     not implemented yet.

     Changed socknal and ranal to use this.

*    Brought ranal up-to-date with new config + mod params for all tunables.
     Ranal gets its local address from the IP of a (single) specified
     interface, or the first "suitable" one found if no interface is specified.

20 years ago* fixed use of %n at end of string in nidstrings.c
eeb [Thu, 19 May 2005 17:53:51 +0000 (17:53 +0000)]
*   fixed use of %n at end of string in nidstrings.c

20 years ago* changed router lookups to always return the local destination
eeb [Thu, 19 May 2005 14:05:33 +0000 (14:05 +0000)]
*   changed router lookups to always return the local destination
    socknal + qswnal changes to match
    simplified "will this be forwarded" predicate in ptl_send()

*   fixed bug in multi-instance socknal to pass the correct NI to the tx
    teardown function to allow for the case where the tx hasn't even had a conn
    assigned yet.

*   fixed routerstat stats formatting

*   removed osc::check_elan_limit() and replaced with max_unfragmented_pages()
    which actually works.  NB it depend on the assumption that file and page
    offsets are aligned!

*   fixed lconf to stay away from portals, except to 'modprobe' for it first in
    'load_module()', and to rmmod all the NALs + portals when ptlrpc gets
    unloaded.

20 years ago* removed any trace of route table file
eeb [Mon, 16 May 2005 07:58:32 +0000 (07:58 +0000)]
*    removed any trace of route table file

*    changed 'routes=' syntax to drop 'route' keyword: syntax is now just a
     list of directives separated by newlines or ';', with comments (#), where
     each directive is <networks> <gateways> [<gateways> ...]

20 years ago* changed route table into a 2-level list; first by target net, then
eeb [Sun, 15 May 2005 18:13:41 +0000 (18:13 +0000)]
*   changed route table into a 2-level list; first by target net, then
    by gateway.

20 years ago* ensure route tables print in same order as constructed
eeb [Sun, 15 May 2005 15:51:09 +0000 (15:51 +0000)]
*   ensure route tables print in same order as constructed

20 years ago* changed 'routes=' syntax...
eeb [Sun, 15 May 2005 12:27:07 +0000 (12:27 +0000)]
*   changed 'routes=' syntax...

       (a) to support pdsh-style range expansion e.g....
           [a,b,c,d-e,f,g,h-i]

       (b) to allow target network expansion so that all the remote networks
           accessible through the same set of gateways can be included in the
           same route command.

20 years ago* multiple TCP networks pass initial tests
eeb [Wed, 11 May 2005 20:12:29 +0000 (20:12 +0000)]
*   multiple TCP networks pass initial tests

20 years ago* lonal part of portals
eeb [Wed, 11 May 2005 16:14:34 +0000 (16:14 +0000)]
*    lonal part of portals

20 years ago* moved lonal into portals module
eeb [Wed, 11 May 2005 12:17:47 +0000 (12:17 +0000)]
*    moved lonal into portals module

20 years ago* lonal moved into portals module
eeb [Wed, 11 May 2005 12:15:44 +0000 (12:15 +0000)]
*   lonal moved into portals module

20 years ago* first cut multiple TCP networks
eeb [Wed, 11 May 2005 06:33:13 +0000 (06:33 +0000)]
*  first cut multiple TCP networks

20 years ago* First cut multiple TCP networks
eeb [Wed, 11 May 2005 06:31:57 +0000 (06:31 +0000)]
*   First cut multiple TCP networks

20 years ago* allow [a,b,c,...] in gateway NID ranges
eeb [Sat, 7 May 2005 18:10:20 +0000 (18:10 +0000)]
*    allow [a,b,c,...] in gateway NID ranges

*    fixed routerstat

20 years ago* changed router to do "obvious" routing automatically if enabled
eeb [Sat, 7 May 2005 03:57:57 +0000 (03:57 +0000)]
*   changed router to do "obvious" routing automatically if enabled

*   tested routing between elan and tcp networks

20 years ago* module loading working with request_module() :))
eeb [Fri, 6 May 2005 16:32:15 +0000 (16:32 +0000)]
*   module loading working with request_module() :))

*   tcp interface enumeration/query working

20 years ago* placeholder (digging out IP interface info)
eeb [Fri, 6 May 2005 14:25:14 +0000 (14:25 +0000)]
*   placeholder (digging out IP interface info)

20 years ago* using request_module() to load the NAL (thanks Andreas)
eeb [Thu, 5 May 2005 17:57:46 +0000 (17:57 +0000)]
*    using request_module() to load the NAL (thanks Andreas)

20 years ago* portals auto-loads required NALs (taken from 'networks=')
eeb [Thu, 5 May 2005 17:12:51 +0000 (17:12 +0000)]
*   portals auto-loads required NALs (taken from 'networks=')

20 years ago* first cut "networks" and "routes" portals module params
eeb [Thu, 5 May 2005 06:43:36 +0000 (06:43 +0000)]
*   first cut "networks" and "routes" portals module params

20 years ago* placeholder
eeb [Thu, 5 May 2005 06:06:00 +0000 (06:06 +0000)]
*   placeholder

20 years ago* placeholder
eeb [Wed, 4 May 2005 09:49:28 +0000 (09:49 +0000)]
*   placeholder

20 years ago* libcfs/nidstrings.c does nid parsing/printing. It's included
eeb [Fri, 29 Apr 2005 19:34:49 +0000 (19:34 +0000)]
*    libcfs/nidstrings.c does nid parsing/printing.  It's included
     in libptlctl.a for userspace utils too.

*    addr@net - style nids working with > 1 local network

*    router part of portals; ptl_send() uses it to chose the NI, but NALs
     still do am-i-there-yet forwarding.  NAL interface changed to match.

*    reworked nal commands into PtlCtl(ni, cmd, arg) and rewired portals ioctl
     handling.  PtlFailNid, now just another PtlCtl.  NAL interface includes
     ctl handler.

*    commented out all lconf portals config, except setting the NID (pending
     socknal working out its own interfaces)

*    removed all portals config from zeroconf and llmount

*    "lctl network" on its own lists all local NIDs (replaces shownid).
     Most (all?) portals config commands still work from lctl.

*    simplified NAL setup/teardown (no separate nal_cmd and router init)

20 years ago* placeholder commit
eeb [Fri, 29 Apr 2005 15:19:45 +0000 (15:19 +0000)]
*    placeholder commit

20 years ago* kernel socknal acceptor; commented out lconf acceptor start/stop
eeb [Tue, 26 Apr 2005 18:28:47 +0000 (18:28 +0000)]
*    kernel socknal acceptor; commented out lconf acceptor start/stop

*    removed socknal's use of sock->file; replaced with atomic refcount in conn

*    moved more socknal tunables into module params/sysfs/sysctl

*    consistent use/naming of socknal refcounting fns

20 years ago* temporarily disabled socknal explicit peer table setup until lconf
eeb [Tue, 26 Apr 2005 02:52:47 +0000 (02:52 +0000)]
*   temporarily disabled socknal explicit peer table setup until lconf
    fixed to stop doing it

*   added socknal support for peer IP == local_address(NID)

*   added CFS_MODULE_PARM macros

*   changed socknal and qswnal tunables to allow setting via module
    parameters

20 years ago* new-config; simple 2 node echo test working with single API interface
eeb [Thu, 21 Apr 2005 17:14:20 +0000 (17:14 +0000)]
*     new-config; simple 2 node echo test working with single API interface

20 years ago* new network config snapshot
eeb [Thu, 21 Apr 2005 00:35:17 +0000 (00:35 +0000)]
*    new network config snapshot
     half-way through lustre uses single NI

20 years agoThis commit was manufactured by cvs2svn to create branch 'b_hd_newconfig'.
cvs2svn [Tue, 19 Apr 2005 23:55:58 +0000 (23:55 +0000)]
This commit was manufactured by cvs2svn to create branch 'b_hd_newconfig'.

20 years agoadd support for sles 8:
jacob [Tue, 19 Apr 2005 23:55:57 +0000 (23:55 +0000)]
add support for sles 8:

 - new target for sles 8 kernel
 - post(un) scripts for kernel rpms
 - fixup quota build on sles 8 and sles 9
 - update patches for a more recent sles 8 kernel

20 years agob=b=6065
jacob [Tue, 19 Apr 2005 23:25:14 +0000 (23:25 +0000)]
b=b=6065
r=eeb
r=adilger

remove noops when stderr isn't present; just fail instead.

20 years agofrom Brian: adjust default lsd location to /usr/sbin instead of /sbin/
ericm [Tue, 19 Apr 2005 22:14:36 +0000 (22:14 +0000)]
from Brian: adjust default lsd location to /usr/sbin instead of /sbin/

20 years ago- added comment in mds_open() and GNS mount points.
yury [Tue, 19 Apr 2005 12:27:07 +0000 (12:27 +0000)]
- added comment in mds_open() and GNS mount points.
- cleanups in sanity-gns.sh
- added one more GNS test - mounting directory with absent mount object

20 years ago- cosmetic changes in mdapi DLD
yury [Tue, 19 Apr 2005 12:08:05 +0000 (12:08 +0000)]
- cosmetic changes in mdapi DLD
- more fixes in GNS stuff:
  - fixes with refcount onto @dchild in ll_gns_mount_object() for error control paths
  - fixing using accessing @dchild in the case of IS_ERR(dchild)

  - disabled checking for GNS file in mds_open(), thus GNS mount points are also
    opening by now. This is needed, because in taking lock we can't reliable to detect
    GNS objects on MDS. Checking just suid bit and skip open such a dirs is not correct
    as one may create structure like dir/dir/dir and mark all of them by suid.

  - cleanups in sanity-gns.sh. Added few new tests for odd conditions like mount object
    is symlink, dir, etc. Added test for changing mount object name to "." and "..".
    Grouping tests into 2 groups:

    (1) general GNS tests - mounting/umounting on timeout with different conditions
    with possible deadlocks, races, etc.

    (2) checking for odd conditions like recursive suid bit marked mount points, etc

20 years ago- more fixes and improvements in GNS stuff:
yury [Mon, 18 Apr 2005 16:37:32 +0000 (16:37 +0000)]
- more fixes and improvements in GNS stuff:
  - small optimization in ll_gns_wait_for_mount(), changing return code semantics to
    usual 0 - ok, < 0 error.

  - handled possible nested mount with returning -ERETSARTSYS and causing endless loop
    of mounting.

  - check if found mount object is regular file to prevent from evil cases like it is
    suid bit marked directory, etc.

  - fix with completting all waiters for dentry mounting in progress.
  - added check for "." and ".." names in setting gns_object_name.
  - fixed default gns_object_name value.

20 years ago- fixed returning @retval in ll_lookup_it() in the case of -ERESTARTSYS
yury [Mon, 18 Apr 2005 12:40:50 +0000 (12:40 +0000)]
- fixed returning @retval in ll_lookup_it() in the case of -ERESTARTSYS

20 years ago- fixed missed comma
yury [Mon, 18 Apr 2005 12:38:33 +0000 (12:38 +0000)]
- fixed missed comma

20 years ago- many fixes in GNS code after Nikita's code review. They are the following:
yury [Mon, 18 Apr 2005 12:13:53 +0000 (12:13 +0000)]
- many fixes in GNS code after Nikita's code review. They are the following:
  - removed ll_d_lookup() as it is not used anymore.

  - removed struts from real_lookup() with calling revalidate() after lookup if dentry
    is marked GNS_MOUNTING. By now all this dentry marking is not needed anymore.

  - ll_gns_mount_object() is now called from ll_revalidate_it() and ll_lookup_it()

  - fixed deadlock when one thread T1 waits for mount completion and another one T2
    created by upcall script waits for dentry->d_inode->i_sem. By now -ERESTARTSYS
    is returned for T2 and others and whren they restart syscall, dentry is mounted
    already or some error occured and GNS state is READY for mount.

  - fixes in comments and header of llite_gns.c

  - fixed possible incorrect spin_unlock() in ll_gns_mount_object()

  - using lookup_one_len() for looking for mount object instead of d_lookup() which
    looks only in dcache.

  - fixed handling of negative mount object dentry in ll_gns_mount_object()

  - fixed readding mount object content. By now kernel_read() is used instead of
    readpage approach.

  - waking up all possible waiters on finishing GNS mount. Nevertheless in current
    stuff nobody seem waits for mount (-RESTARTSYS is used instead), completion
    exists and it is semanticaly correct to wake up all its waiters. In fact this
    is needed for possible future using. If it happen that nobody will wait on it
    and it is counted not needed, it will be removed.

20 years agofrom Alex: temporarily fix for mds hanging on mountain: spinlock might
ericm [Fri, 15 Apr 2005 22:22:57 +0000 (22:22 +0000)]
from Alex: temporarily fix for mds hanging on mountain: spinlock might
be called in interrupt context.

20 years agosmfs support in basic tests
tappro [Wed, 13 Apr 2005 17:14:15 +0000 (17:14 +0000)]
smfs support in basic tests
MDS_BACKFSTYPE and OST_BACKFSTYPE are defined for this.
FSTYPE is ldiskfs for 2.6 kernel by default