glusterfs.git/rpc/rpc-transport/rdma/src/rdma.h, branch v6.3

Revert "iobuf: Get rid of pre allocated iobuf_pool and use per thread mem pool"

2019-01-08T11:16:03+00:00

This reverts commit b87c397091bac6a4a6dec4e45a7671fad4a11770.

There seems to be some performance regression with the patch and hence recommended to have it reverted.

Updates: #325
Change-Id: Id85d6203173a44fad6cf51d39b3e96f37afcec09

iobuf: Get rid of pre allocated iobuf_pool and use per thread mem pool

2018-12-18T09:35:24+00:00

The current implementation of iobuf_pool has two problems:
- prealloc of 12.5MB memory, this limits the scale factor of the gluster
  processes due to RAM requirements
- lock contention, as the current implementation has one global
  iobuf_pool lock. Credits for debugging and addressing the same goes to
  Krutika Dhananjay . Issue: #410

Hence changing the iobuf implementation to use per thread mem pool.
This may theoritically appear to cause perf dip as there is no preallocation.
But per thread mem pool will not have significant perf impact as the last
allocated memory is kept alive for subsequent allocs, for some time.
The worst case would be if iobufs requested are of random sizes each time.
The best case is, if we get iobuf request of the same size. From the perf
tests, this patch did not seem to cause any perf decrease.

Note that, with this patch, the rdma performance is going to degrade
drastically. In one of the previous patchsets we had fixes to not
degrade rdma perf, but rdma is not supported and also not tested [1].
Hence the decision was to not have code in rdma that is not tested
and not supported.

[1] https://lists.gluster.org/pipermail/gluster-users.old/2018-July/034400.html

Updates: #325
Change-Id: Ic2ef3bd498f9250dea25f25ba0c01fde19584b27
Signed-off-by: Poornima G

libglusterfs: Move devel headers under glusterfs directory

2018-12-05T21:47:04+00:00

libglusterfs devel package headers are referenced in code using
include semantics for a program, this while it works can be better
especially when dealing with out of tree xlator builds or in
general out of tree devel package usage.

Towards this, the following changes are done,
- moved all devel headers under a glusterfs directory
- Included these headers using system header notation <> in all
code outside of libglusterfs
- Included these headers using own program notation "" within
libglusterfs

This change although big, is just moving around the headers and
making it correct when including these headers from other sources.

This helps us correctly include libglusterfs includes without
namespace conflicts.

Change-Id: Id2a98854e671a7ee5d73be44da5ba1a74252423b
Updates: bz#1193929
Signed-off-by: ShyamsundarR

Land clang-format changes

2018-09-12T11:52:48+00:00

Change-Id: I6f5d8140a06f3c1b2d196849299f8d483028d33b

build: rename event.h to gf-event.h

2018-07-27T17:37:32+00:00

Newer FreeBSD versions (noticed with 10.3-RELEASE) provide a event.h
file that on occasion gets included instead of the libglusterfs file.
When this happens, 'struct event_pool' will not be defined and building
will fail with errors like:

    autoscale-threads.c:18:55: error: incomplete definition of type 'struct event_pool'
            int                      thread_count   = pool->eventthreadcount;
                                                      ~~~~^
    autoscale-threads.c:17:16: note: forward declaration of 'struct event_pool'
            struct event_pool       *pool           = ctx->event_pool;
                   ^

This problem is caused by 'pkg-config --cflags uuid' that adds
/usr/local/include to the GF_CPPFLAGS. The use of libuuid is preferred
so that the contrib/uuid/ directory can be removed.

By renaming event.h to gf-event.h there is no conflict between the
different event.h files anymore and compiling on FreeBSD works without
issues.

Change-Id: Ie69f6b8a4f8f8e9630d39a86693eb74674f0f763
Updates: bz#1607319
Signed-off-by: Niels de Vos

rpc-transport/rdma: Add a mutex for the list of RDMA Memory Region(MR) access

2017-12-12T05:49:35+00:00

Problem: gf_rdma_device_t->all_mr is a __gf_rdma_arena_mr(includes MR content)
	 kind of list in the rdma rpc-transport. The rdma rpc-transport will
	 add/delete items to the list when MRs register, deregister, and free.
	 Because gf_rdma_device_t->all_mr is used by different threads
	 and it is not mutex protected, rdma transport maybe access obsolete
	 items in it.

Solution: Add a mutex protection for the gf_rdma_device_t->all_mr.

Change-Id: I2b7de0f7aa516b90bb6f3c6aae3aadd23b243900
BUG: 1522651
Signed-off-by: Yi Wang

protocol/server: make listen backlog value as configurable

2017-06-08T15:32:30+00:00

problem:

When we call listen from protocol/server, we are giving a
hard coded valie of 10 if it is not manually given.
With multiplexing, especially when glusterd restarts all
clients may try to connect to the server at a time.
Which will result in overflowing the queue, and kernel
will complain about the errors.

Solution:

This patch will introduce a volume set command to make backlog
value as a configurable. This patch also changes the default
values for backlog from 10 to 128. This changes is only applicable
for sockets listening from protocol.

Example:

gluster volume set  transport.listen-backlog 1024

Note: 1 Brick has to be restarted to get this value in effect
      2 This changes won't be reflected in glusterd, or other
        xlators which calls listen. If you need, you have to
        add this option to the volfile.

Change-Id: I0c5a2bbf28b5db612f9979e7560e05dd82b41477
BUG: 1456405
Signed-off-by: Mohammed Rafi KC 
Reviewed-on: https://review.gluster.org/17411
Smoke: Gluster Build System 
NetBSD-regression: NetBSD Build System 
CentOS-regression: Gluster Build System 
Reviewed-by: Raghavendra G 
Reviewed-by: Raghavendra Talur 
Reviewed-by: Atin Mukherjee 
Reviewed-by: Niels de Vos 
Reviewed-by: Jeff Darcy

build: do not #include "config.h" in each file

2015-05-29T12:43:17+00:00

Instead of including config.h in each file, and have the additional
config.h included from the compiler commandline (-include option).

When a .c file tests for a certain #define, and config.h was not
included, incorrect assumtions were made. With this change, it can not
happen again.

BUG: 1222319
Change-Id: I4f9097b8740b81ecfe8b218d52ca50361f74cb64
Signed-off-by: Niels de Vos 
Reviewed-on: http://review.gluster.org/10808
Tested-by: Gluster Build System 
Tested-by: NetBSD Build System
Reviewed-by: Kaleb KEITHLEY 
Reviewed-by: Pranith Kumar Karampuri

rdma:properly handle iobuf_pool when rdma transport is unloaded

2015-05-05T13:15:16+00:00

We are registering iobuf_pool with rdma. When rdma transport is
unloaded, we need to deregister all the buffers registered with
rdma. Otherwise iobuf_arena destroy will fail.

Also if rdma.so is loaded again, then register iobuf_pool with
rdma

Change-Id: Ic197721a44ba11dce41e03058e0a73901248c541
BUG: 1200704
Signed-off-by: Mohammed Rafi KC 
Reviewed-on: http://review.gluster.org/9854
Tested-by: NetBSD Build System
Tested-by: Gluster Build System 
Reviewed-by: Raghavendra Talur

rdma: Free resources related to iobuf in fini

2015-03-11T02:28:29+00:00

If rdma transport is destroyed because of any reason,
then rdma.so will be unloaded. But we are not setting
iobuf registeration function to null. After this, if
an iobuf request is came, then we will try to call a
function which is not loaded.

Change-Id: I3293f9974e16d8e865131785ee697ea02be8cdfc
BUG: 1187456
Signed-off-by: Mohammed Rafi KC 
Reviewed-on: http://review.gluster.org/9697
Tested-by: Gluster Build System 
Reviewed-by: Raghavendra Talur 
Reviewed-by: Vijay Bellur