In 2.10 the entire set of changes to using ktime function is not in
creating a situation where routers consider peers dead, essentially
breaking routing functionality.

Revert the changes that were made to LNet as part of LU-6245.

Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: Id0765e8332a8b97167d3a602e6410f5bb6a48137
Reviewed-on: https://review.whamcloud.com/32082


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Sebastien Buisson <sbuisson@ddn.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

It was found for the case of routers which have LNet enabled
ethernet and infiniband that they couldn't communicate properly
with other. This was due to ko2iblnd still using jiffies and
the socklnd driver using time64_t. After some discussion it was
deceided that the best thing to do is roll back instead of
additional changes.

Revert "LU-9397 ksocklnd: move remaining time handling to 64 bits"

This reverts commit 59c25356.

Change-Id: I42fd5620ab8131cd5fdf4bf0ef15553f6eabe550
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/32015


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Sebastien Buisson <sbuisson@ddn.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

In cfs_ip_min_max(), (nidrange->nr_all == 1) means this nid range
is a full IP address range(*.*.*.*). In this case, we don't need
to compare it to any other nid range, but set min_nid to 0.0.0.0
and max_nid to 255.255.255.255 directly.

Also, test_10d is added to sanity-sec.sh to verify this patch and
some code cleanup is done for jt_nodemap_add/del_range().

Change minimum MGS version to 2.10.1

Lustre-change: https://review.whamcloud.com/31684


Lustre-commit: 23026632

Change-Id: I72c546b060f9e123204a566a3bd373b4f017502d
Signed-off-by: Emoly Liu <emoly.liu@intel.com>
Reviewed-by: Sebastien Buisson <sbuisson@ddn.com>
Reviewed-by: Fan Yong <fan.yong@intel.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/31950


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

This patch fixes the contiguous range check to allow the addition of
multiple "full" ([0-255]) ranges. As part of this change,
is_contiguous and find_min_max are combined as they were always
called together and the logic is fairly similar. This also removes
the multiple range expression support, since it was broken.

Also, sanity-sec.sh test_10c is added to verify this patch.

Lustre-change: https://review.whamcloud.com/24397


Lustre-commit: eac95a65

Signed-off-by: Kit Westneat <kit.westneat@gmail.com>
Signed-off-by: Emoly Liu <emoly.liu@intel.com>
Change-Id: I3c49a077039327fcbde87196f82db140f67a74d0
Reviewed-by: Sebastien Buisson <sbuisson@ddn.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/31521


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

list_add was being used erroneously. The logic should be to move
the txs on ibp_tx_queue on a local list which is then processed.
The code, however, did the reverse, which would result in the
pending txs not processed and thus dropped silently. This in turn
would lead to peers reference counts at the LNet layer not
decremented since lnet_finalize() might not be called for a message.

Initialize local list and use list_splice_init() to move
transmits on the ibp_tx_queue to the local list.

Lustre-change: https://review.whamcloud.com/31374


Lustre-commit: f5c6228f

Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I6b36f709db2c89e53e0b3354883a8a1b1052a1dd
Reviewed-by: Doug Oucharek <dougso@me.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/31520


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

Request dynamic minor allocation when registering /dev/lnet and
/dev/obd. Remove the obsolete create-device-if-not-found code from
register_ioc_dev().

Lustre-change: https://review.whamcloud.com/29741


Lustre-commit: e446c166

Signed-off-by: John L. Hammond <john.hammond@intel.com>
Change-Id: I59c70912b4729f58a76dc6107b3e1d7379c6d7a3
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Jian Yu <jian.yu@intel.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/29945


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>

cmid will be destroyed at OFED if kiblnd_cm_callback return error.
if error happen before the end of kiblnd_connect_peer, it will touch
destroyed cmid and fail as
(o2iblnd_cb.c:1315:kiblnd_connect_peer())
            ASSERTION( cmid->device != ((void *)0) ) failed:

Lustre-change: https://review.whamcloud.com/29134


Lustre-commit: 576551cb

Seagate-bug-id: MRP-4592
Signed-off-by: Alexander Boyko <alexander.boyko@seagate.com>
Change-Id: I83eb5bceeb567acef0316498b936d25d6c6ccd95
Reviewed-by: Alexey Lyashkov <c17817@cray.com>
Reviewed-by: Doug Oucharek <dougso@me.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/29881


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

The arguments args->lstio_ses_force and args->lstio_ses_timeout are
in the incorrect order. Fix this by swapping them around.

Detected by CoverityScan, CID#1226833 ("Arguments in wrong order")

Test-Parameters: trivial testlist=lnet-selftest

Lustre-change: https://review.whamcloud.com/28487


Lustre-commit: 0a80067c

Change-Id: If11c574655425db5bbf21ba2264be8d83a7e8bf8
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/28763


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

During the development of the linux 4.11 kernel it was discovered
that the kernel socket layer could get into lockdep situation. To
handle this a new bool argument was added to the accept member
of struct socket. For LNet we can always pass false.

Lustre-commit: 15045c90
Lustre-change: https://review.whamcloud.com/27642



Change-Id: I420cda95b70cf927b1a6e3493b631bc5a3585d74
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/27642


Reviewed-by: Doug Oucharek <doug@cadentcomputing.com>
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
Reviewed-on: https://review.whamcloud.com/28463


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

When tx credits are returned if there are pending messages they
need to be sent. Messages could have different tx_cpts, so the
correct one needs to be locked. After lnet_post_send_locked(),
if we locked a different CPT then we need to relock the correct one
However, as part of lnet_post_send_locked(), lnet_finalze() can
be called which can free the message. Therefore, the cpt of the
message being passed must be cached in order to prevent access to
freed memory.

Lustre-change: https://review.whamcloud.com/28308


Lustre-commit: 7af6307b

Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I959fdc30daf87b5575d8371da20d5cf6f64e7d3c
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/28439


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

The intent of this function is to get the cpt nearest to the
memory described by the MD.

There are three scenarios that must be handled:
1. The memory is described by an lnet_kiov_t structure
 -> this describes kernel pages
2. The memory is described by a struct kvec
 -> this describes kernel logical addresses
3. The memory is a contiguous buffer allocated via vmalloc

For case 1 and 2 we look at the first vector which contains
the data to be DMAed, taking into consideration the msg offset.

For case 2 we have to take the extra step of translating the kernel
logical address to a physical page using virt_to_page() macro.

For case 3 we need to use is_vmalloc_addr() and vmalloc_to_page to
get the associated page to be able to identify the CPT.

o2iblnd uses the same strategy when it's mapping the memory into
a scatter/gather list. Therefore, lnet_kvaddr_to_page() common
function was created to be used by both the o2iblnd and
lnet_cpt_of_md()

kmap_to_page() performs the high memory check which
lnet_kvaddr_to_page() does. However, unlike the latter it handles
the highmem case properly instead of calling LBUG. It's not
100% clear why the code was written that way. Since the legacy
code will need to still be maintained, adding kmap_to_page() will
not simplify the code. Furthermore, the behavior for kernels
which export kmap_to_page() will be different from kernels which
do not. At worst calling kmap_to_page() might mask some problems
which would've been caught by the LBUG earlier on. However, at
the time of this fix, that LBUG has never been observed.

Lustre-change: https://review.whamcloud.com/28165


Lustre-commit: 43b0e632

Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I2c67e5df77d60112bf27f900e0325d189f193aed
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/28400


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

When the obdfilter code was split into separate OFD and OSD modules,
the bulk IO page allocation was implemented to use GFP_NOFS to avoid
allocations recursing into the filesystem and causing deadlocks.

However, this is only possible if the RPC is coming from a local
client, as we might end up waiting on a page sent in the request we're
serving. Local RPCs use __GFP_HIGHMEM so that the pages can use all of
the available memory on the OSS on 32-bit machines.

It is possible to use more aggressive GFP_HIGHUSER flags for non-local
clients to be able to generate more memory pressure on the OSS and
allow inactive pages to be reclaimed, since the OSS doesn't have any
other processes or allocations that generate memory reclaim pressure.

See also b=17576 (bdf50dc9) and b=19529 (3dcf18d3) for details.

The patch also implements an LNet function to determine if a client NID
is local or not.  This becomes more complex in the LNet Multi-Rail world
and it is really LNet's job to handle NIDs, not that of Lustre.

Lustre-change: https://review.whamcloud.com/27908


Lustre-commit: b0ab95d6

Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Change-Id: I2806c9c5c2fe269669eafdafaf2310924c3ebbe5
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Patrick Farrell <paf@cray.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/28318


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

Trying to page align the remote_addr for IB_RDMA_WRITE work
requests is triggering "dump_cqe" errors from MOFED 4.x + mlx5.

This patch removes the address masking we were doing with FastReg
which was trying to page align remote_addr values. I am also
removing the setting of "mr->iova" with FastReg as this is being
done in the call to ib_map_mr_sg() and could cause problems.

Lustre-change: https://review.whamcloud.com/27149


Lustre-commit: 6c634180

Signed-off-by: Doug Oucharek <doug.s.oucharek@intel.com>
Change-Id: If35baa467d8d60866f709b5feea7f619063c6da4
Reviewed-by: Gu Zheng <gzheng@ddn.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/28237


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

Make sure to unlock the api mutex properly
in lnet_dyn_add_net()

Lustre-change: https://review.whamcloud.com/27907


Lustre-commit: 65326ab2

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I786545de690ea5966771be3e84d3561b794d55ec
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Signed-off-by: Minh Diep <minh.diep@intel.com>
Reviewed-on: https://review.whamcloud.com/28236


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: John L. Hammond <john.hammond@intel.com>

Unlock lnet_net_lock in case of error in function lnet_select_pathway().

Change-Id: Ib48fb3aebdc60bafff80f5c52b90301830ca4afa
Signed-off-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-on: https://review.whamcloud.com/27455


Tested-by: Jenkins
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Rework CPU partition code in the way of make it more tolerant to
offline CPUs and empty nodes.

For example, in KNL:
available: 8 nodes (0-7)
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221
node 0 size: 24472 MB
node 0 free: 12409 MB
node 1 cpus: 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239
node 1 size: 24576 MB
node 1 free: 20388 MB
node 2 cpus: 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
node 2 size: 24576 MB
node 2 free: 20621 MB
node 3 cpus: 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271
node 3 size: 24576 MB
node 3 free: 21183 MB
node 4 cpus:
node 4 size: 4096 MB
node 4 free: 3982 MB
node 5 cpus:
node 5 size: 4096 MB
node 5 free: 3982 MB
node 6 cpus:
node 6 size: 4096 MB
node 6 free: 3982 MB
node 7 cpus:
node 7 size: 4096 MB
node 7 free: 3981 MB
node distances:
node   0   1   2   3   4   5   6   7
  0:  10  21  21  21  31  41  41  41
  1:  21  10  21  21  41  31  41  41
  2:  21  21  10  21  41  41  31  41
  3:  21  21  21  10  41  41  41  31
  4:  31  41  41  41  10  41  41  41
  5:  41  31  41  41  41  10  41  41
  6:  41  41  31  41  41  41  10  41
  7:  41  41  41  31  41  41  41  10

Contain the fix for LU-8492 ptlrpc: Correctly calculate hrp->hrp_nthrs
Fix an error code return which was introduced in
commit def25e9c

Change-Id: I7f64a20ee009a88e836f592ce044400f07ffbcdd
Signed-off-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-on: https://review.whamcloud.com/23222


Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Increment the per NI stats for messages being routed.
This will give a better view of the traffic distribution
over multiple peer interfaces.

Added extra trace messages to track the messages
sent and received.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I213b1b36e9787d25705d91d091ad9e9c6a5b2ae8
Reviewed-on: https://review.whamcloud.com/26907


Tested-by: Jenkins
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Make sure to set all NIs to the proper LND tunables
specified.

Add ntx tunable to dynamic configuration. This way all
tunables required to tune OPA performance can be configured
via lnetctl, allowing the ability to tune OPA network and IB
network differently

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I015f3f959bd46784d4607bd4259b4640303dc362
Reviewed-on: https://review.whamcloud.com/27263


Tested-by: Jenkins
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

lnet_peer_primary_nid() is called from lnet_parse. It checks
ln_state outside the net lock, causing a race condition
during shutdown where the code expects the state to be
running, but it's stopping or shutdown.

Fixed the issue by renaming lnet_peer_primary_nid() to
lnet_peer_primary_nid_locked(). This function is now called
when lnet_net_lock is held in lnet_parse().

In lnet_create_reply_msg() we already have access to the
msg_txpeer, so we lookup the primary_nid directly

Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I0518cdbec95b38bd8690517320b601676ae259f0
Reviewed-on: https://review.whamcloud.com/27262


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Examination of the ksocklnd time handle revealed that the
code only requires second level precision. Since this is
the case we can move away from using jiffies to time64_t.
This allows us to be independent of the HZ settings in
addition to making it clear what is time handling, using
time64_t verses unsigned long. In the process we remove
many of the various libcfs time wrappers as well.

Change-Id: I968630ef94febd4bff703fb633e677996939f95b
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/26813


Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Make the LND code handle empty CPTs.

If a scheduler is associated with an empty CPT it will have no
threads created. If a NID hashes to that CPT, then pick the
next scheduler which does have at least 1 started thread.
Associate the connection with the CPT of the selected scheduler

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I8ed83f295fe9852537d4bb063a4d8271c6a45c2c
Reviewed-on: https://review.whamcloud.com/27145


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

The CFS_TIME_T macro serves no real purpose as we stopped using time_t
and changed over to time64_t, so we can remove the last remaining uses
of this.

Two uses of this macro are incorrect and refer to jiffies values
rather than time_t, and one refers to an inode timespec that gets
changed separately.

Linux-commit: 93d3a405a168fba4450bdda793149e3cd4174736

Change-Id: I548ec8fffc9c46b8b2025b094f1e5d9cd469e3b7
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Oleg Drokin <green@linuxhacker.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/27025


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Mike Pershin <mike.pershin@intel.com>

LU-8943 activated the ability to have multiple connections
between peers.  If any of those connections need to be
reconnected (i.e. parameter renegotiation), we were getting
an assert from kiblnd_reconnect_peer() which was not changed
to allow for having multiple connections ongoing at the
same time.

This patch gets rid of the assert from kiblnd_reconnect_peer()
which is no longer valid after LU-8943.

Signed-off-by: Doug Oucharek <doug.s.oucharek@intel.com>
Change-Id: I9cc7fae8836f2648603018fac38a88e3f90ec190
Reviewed-on: https://review.whamcloud.com/27139


Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Tested-by: Jenkins
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

The cfs_duration_usec() function has a timeval as its output, which we
want to avoid in general because of the y2038 problem.

There are only two locations remaining in lustre, so we can for now
eplace one with jiffies_to_timeval(), which is a generic kernel function
that does the same thing, the other can just use jiffies_to_usecs()
and completely avoid the timeval.

This is not a full solution yet, but it's a small step that lets us
build a larger portion of lustre without this reference to timeval in
a header file, and avoid triggering automated checking tools that wants
to warn about timeval.

Linux-commit: 70513c5d17b9812cc218e8b4c7826ebb5f375d9a

Test-Parameters: trivial testlist=lnet-selftest

Change-Id: If39f4d4857a2b3210bb0dc634b8bb42530df83dc
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Oleg Drokin <green@linuxhacker.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/27019


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

The FastReg support in ko2iblnd was not unmapping pool items
causing the items to leak.  In addition, the mapping code
is not growing the pool like we do with FMR.

This patch makes sure we are unmapping FastReg pool elements
when we are done with them.  It also makes sure the pool
will grow when we depleat the pool.

Signed-off-by: Doug Oucharek <doug.s.oucharek@intel.com>
Change-Id: I4b4ba4de72941b38c4115a00a992cfd1e78e9e49
Reviewed-on: https://review.whamcloud.com/27015


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andrew Perepechko <andrew.perepechko@seagate.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Deleting peers with lnetctl command
"import --del < config.yaml" throws error

The above command deletes prim_nid first
and then tries deleting the other nids which
results in error, since deleting the primary_nid
deletes the entire peer and then after that we try
to delete non-existent NIDs.

The behavior should be if the primary_nid is
present in the list of NIDs then delete the
entire peer, otherwise delete only the NIDs
specified within the peer

Test-Parameters: trivial
Signed-off-by: Sonia Sharma <sonia.sharma@intel.com>
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I55114fca4d332c950872bd446e02e4f0904ee716
Reviewed-on: https://review.whamcloud.com/27001


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Currently, the fix in LU-5718 which allows for multiple sges
to deal with RDMA fragmentation is turned off by deafult
(set to 1).  This patch changes the default to 2 so
RDMA fragmentation is fixed by default.

Test-Parameters: trivial
Signed-off-by: Doug Oucharek <doug.s.oucharek@intel.com>
Change-Id: I8a29a7b32ababd37cbc471664083362bc7253d97
Reviewed-on: https://review.whamcloud.com/26911


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Ned Bass <bass6@llnl.gov>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Changing all calls in the ksocklnd from sock_create() to
sock_create_kern().

Signed-off-by: Doug Oucharek <doug.s.oucharek@intel.com>
Change-Id: Ib8b175e73478b1edfb5e8cd3491e589e8267f52a
Reviewed-on: https://review.whamcloud.com/26958


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Olaf Weber <olaf.weber@hpe.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

So that it can be included in other projects.

Signed-off-by: Bobi Jam <bobijam.xu@intel.com>
Change-Id: I980578742fd194e2464870f1ab8d6a9ae8deb9e2
Reviewed-on: https://review.whamcloud.com/26859


Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Niu Yawei <yawei.niu@intel.com>
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

OPA driver optimizations are based on the MPI model where it is
expected to have multiple endpoints between two given nodes. To
enable this optimization for Lustre, we need to make it possible,
via an LND-specific tuneable, to create multiple endpoints and to
balance the traffic over them.

Both sides of a connection must have this patch for it to work.
Only the active side of the connection (usually the client)
needs to have the new tuneable set > 1.

Signed-off-by: Doug Oucharek <doug.s.oucharek@intel.com>
Change-Id: Iaf3b49bf0aecf79cb67eb1bacba1940cd811b2fb
Reviewed-on: https://review.whamcloud.com/25168


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Reviewed-by: Dmitry Eremin <dmitry.eremin@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

A flags argument was added to ib_alloc_pd() in Linux 4.9 commit
ed082d36a7b2c27d1cda55fdfb28af18040c4a89. The fix for LU-9026, Lustre
commit e4297ef3, accounted for this
change by checking for the removal of ib_get_dma_mr() (which happened
separately). However, SLES 12 SP3 beta 1 adopted the extra argument
to ib_alloc_pd(), but retains the ib_get_dma_mr() function. As a
result, we need an explicit check for the two argument version of
ib_alloc_pd().

Signed-off-by: Chris Horn <hornc@cray.com>
Change-Id: Iecde347e9f18149cac63e243082a2686de260ba7
Reviewed-on: https://review.whamcloud.com/26934


Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Tested-by: Jenkins
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

selftest always responded to the primary nid of the peer rather than
the source of the message, which it should be.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Signed-off-by: Olaf Weber <olaf@sgi.com>
Change-Id: I14a4b6ffc5882cb23298429d8a4bd0bcb0a8a5be
Reviewed-on: https://review.whamcloud.com/26723


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

LNET_MAX_INTERFACES is the number of interfaces supported by
interface bonding in the ksocknal LND. It shows up in LNet
because a number of data structures are shared between LNDs.

Rename it to LNET_NUM_INTERFACES to reduce the confusion of
what it does.

Test-Parameters: trivial
Signed-off-by: Olaf Weber <olaf@sgi.com>
Change-Id: Ibc1d85a379d6616eb1db2fcb54aaffc835ffa9f4
Reviewed-on: https://review.whamcloud.com/26693


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

In lnet_select_pathway() sending to the loopback NID is handled
as a special case, because there are no credits involved. (The
loopback NID doesn't use credits, and therefore does not have
any credits. If a message goes through the credit-managing code
it therefore ends up waiting indefinitely for credits to become
available.)

The check whether we're sending over the loopback NID must be
done after we've completed choosing the NI to send over. In its
present location it only handles the case where the loopback
NID was explicitly passed in as the source NID.

(Lustre does not exercise this code path during normal operation,
the bug was encountered while testing code for the peer discovery
feature.)

Test-Parameters: trivial
Signed-off-by: Olaf Weber <olaf@sgi.com>
Change-Id: Ifa25abf508214ae363a2f1bb04ffeab1891a2564
Reviewed-on: https://review.whamcloud.com/26692


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

When an attempt to send a message fails, for example because no
connection could be established with the remote address, socklnd
drops the message. For a PUT or REPLY message with non-zero
payload, ksocknal_tx_done() calls lnet_finalize() with -EIO
as the error code. But for an ACK or GET message there is no
payload, and lnet_finalize() is called with 0 (no error) as the
error code. This leaves upper layers to rely on other means to
determine that sending the message did actually fail, and that
(for example) no REPLY will ever answer a failed GET.

Add an error code parameter to ksocknal_tx_done().

In ksocknal_txlist_done() change the 0/1 'error' indicator to be
an actual error code that is passed on the ksocknal_tx_done().
Update the callers of ksocknal_txlist_done() to pass in the error
code if they have encountered an error.

Test-Parameters: trivial
Signed-off-by: Olaf Weber <olaf@sgi.com>
Change-Id: I66b897a31e537e70dcc2622ffdfcc6e96fa93193
Reviewed-on: https://review.whamcloud.com/26691


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

The locking changes for the lnet_net_lock made for Multi-Rail
introduce a race in the LNet shutdown path. The code keeps two
states in the_lnet.ln_shutdown: 0 means LNet is either up and
running or shut down, while 1 means lnet is shutting down. In
lnet_select_pathway() if we need to restart and drop and relock
the lnet_net_lock we can find that LNet went from running to
stopped, and not be able to tell the difference.

Replace ln_shutdown with a three-state ln_state patterned on
ln_rc_state: states are LNET_STATE_SHUTDOWN, LNET_STATE_RUNNING,
and LNET_STATE_STOPPING. Most checks against ln_shutdown now test
ln_state against LNET_STATE_RUNNING. LNet moves to RUNNING state
in lnet_startup_lndnets().

Test-Parameters: trivial
Signed-off-by: Olaf Weber <olaf@sgi.com>
Change-Id: I7afcbeb793dfa4d0a361e421ae06a99b7d4db903
Reviewed-on: https://review.whamcloud.com/26690


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

When the patch for LU-5710 landed, a check for message size was landed
that should not have been.  This check was part of a patch in LU-7650
which was later pulled because it broke things. LU-5718 picked up this
code via its many rebases (it too forever to land LU-5718 which is the
core problem here).

This patch removes that messaage size check.

Test-Parameters: trivial
Signed-off-by: Doug Oucharek <doug.s.oucharek@intel.com>
Change-Id: I3d114ec16cfbfd994efd9aee55e28a09159597be
Reviewed-on: https://review.whamcloud.com/26891


Tested-by: Jenkins
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Sonia Sharma <sonia.sharma@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Amir Shehata <amir.shehata@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

To avoid backwards compatibility issues between base MR
and Dynamic Discovery standardize the ioctl interface by
bringing in changes to the interface required by
Dynamic Discovery now.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I014d74943b893ec24e3d42e1eb6824d755460c2b
Reviewed-on: https://review.whamcloud.com/26689


Tested-by: Jenkins
Reviewed-by: Olaf Weber <olaf@sgi.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Remove a debug ioctl that was added to allow for debug
messages from user space. However, the code is currently
not being used.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Signed-off-by: Olaf Weber <olaf@sgi.com>
Change-Id: Ifd2bee73ef507bd07296af76dac1caf08ded9e64
Reviewed-on: https://review.whamcloud.com/26688


Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

Fixed a set of issues found while running static analysis.

Test-Parameters: trivial
Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Change-Id: I22ddfdda86c979c7a300ab9df777efbdd5973ac5
Reviewed-on: https://review.whamcloud.com/26687


Tested-by: Jenkins
Reviewed-by: Olaf Weber <olaf@sgi.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Doug Oucharek <doug.s.oucharek@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>