From: Chuck Lever <chuck.lever@oracle.com>
To: Anna.Schumaker@netapp.com
Cc: linux-rdma@vger.kernel.org, linux-nfs@vger.kernel.org
Subject: [PATCH v4 17/21] xprtrdma: Allocate each struct rpcrdma_mw separately
Date: Tue, 22 Jul 2014 10:53:27 -0400 [thread overview]
Message-ID: <20140722145327.6010.85935.stgit@manet.1015granger.net> (raw)
In-Reply-To: <20140722144459.6010.99389.stgit@manet.1015granger.net>
Currently rpcrdma_buffer_create() allocates struct rpcrdma_mw's as
a single contiguous area of memory. It amounts to quite a bit of
memory, and there's no requirement for these to be carved from a
single piece of contiguous memory.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
---
net/sunrpc/xprtrdma/verbs.c | 242 +++++++++++++++++++++++++------------------
1 file changed, 143 insertions(+), 99 deletions(-)
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 1294338..f7691be 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -1005,9 +1005,91 @@ rpcrdma_ep_disconnect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
return rc;
}
-/*
- * Initialize buffer memory
- */
+static int
+rpcrdma_init_fmrs(struct rpcrdma_ia *ia, struct rpcrdma_buffer *buf)
+{
+ int mr_access_flags = IB_ACCESS_REMOTE_WRITE | IB_ACCESS_REMOTE_READ;
+ struct ib_fmr_attr fmr_attr = {
+ .max_pages = RPCRDMA_MAX_DATA_SEGS,
+ .max_maps = 1,
+ .page_shift = PAGE_SHIFT
+ };
+ struct rpcrdma_mw *r;
+ int i, rc;
+
+ i = (buf->rb_max_requests + 1) * RPCRDMA_MAX_SEGS;
+ dprintk("RPC: %s: initalizing %d FMRs\n", __func__, i);
+
+ while (i--) {
+ r = kzalloc(sizeof(*r), GFP_KERNEL);
+ if (r == NULL)
+ return -ENOMEM;
+
+ r->r.fmr = ib_alloc_fmr(ia->ri_pd, mr_access_flags, &fmr_attr);
+ if (IS_ERR(r->r.fmr)) {
+ rc = PTR_ERR(r->r.fmr);
+ dprintk("RPC: %s: ib_alloc_fmr failed %i\n",
+ __func__, rc);
+ goto out_free;
+ }
+
+ list_add(&r->mw_list, &buf->rb_mws);
+ list_add(&r->mw_all, &buf->rb_all);
+ }
+ return 0;
+
+out_free:
+ kfree(r);
+ return rc;
+}
+
+static int
+rpcrdma_init_frmrs(struct rpcrdma_ia *ia, struct rpcrdma_buffer *buf)
+{
+ struct rpcrdma_frmr *f;
+ struct rpcrdma_mw *r;
+ int i, rc;
+
+ i = (buf->rb_max_requests + 1) * RPCRDMA_MAX_SEGS;
+ dprintk("RPC: %s: initalizing %d FRMRs\n", __func__, i);
+
+ while (i--) {
+ r = kzalloc(sizeof(*r), GFP_KERNEL);
+ if (r == NULL)
+ return -ENOMEM;
+ f = &r->r.frmr;
+
+ f->fr_mr = ib_alloc_fast_reg_mr(ia->ri_pd,
+ ia->ri_max_frmr_depth);
+ if (IS_ERR(f->fr_mr)) {
+ rc = PTR_ERR(f->fr_mr);
+ dprintk("RPC: %s: ib_alloc_fast_reg_mr "
+ "failed %i\n", __func__, rc);
+ goto out_free;
+ }
+
+ f->fr_pgl = ib_alloc_fast_reg_page_list(ia->ri_id->device,
+ ia->ri_max_frmr_depth);
+ if (IS_ERR(f->fr_pgl)) {
+ rc = PTR_ERR(f->fr_pgl);
+ dprintk("RPC: %s: ib_alloc_fast_reg_page_list "
+ "failed %i\n", __func__, rc);
+
+ ib_dereg_mr(f->fr_mr);
+ goto out_free;
+ }
+
+ list_add(&r->mw_list, &buf->rb_mws);
+ list_add(&r->mw_all, &buf->rb_all);
+ }
+
+ return 0;
+
+out_free:
+ kfree(r);
+ return rc;
+}
+
int
rpcrdma_buffer_create(struct rpcrdma_buffer *buf, struct rpcrdma_ep *ep,
struct rpcrdma_ia *ia, struct rpcrdma_create_data_internal *cdata)
@@ -1015,7 +1097,6 @@ rpcrdma_buffer_create(struct rpcrdma_buffer *buf, struct rpcrdma_ep *ep,
char *p;
size_t len, rlen, wlen;
int i, rc;
- struct rpcrdma_mw *r;
buf->rb_max_requests = cdata->max_requests;
spin_lock_init(&buf->rb_lock);
@@ -1026,28 +1107,12 @@ rpcrdma_buffer_create(struct rpcrdma_buffer *buf, struct rpcrdma_ep *ep,
* 2. arrays of struct rpcrdma_req to fill in pointers
* 3. array of struct rpcrdma_rep for replies
* 4. padding, if any
- * 5. mw's, fmr's or frmr's, if any
* Send/recv buffers in req/rep need to be registered
*/
-
len = buf->rb_max_requests *
(sizeof(struct rpcrdma_req *) + sizeof(struct rpcrdma_rep *));
len += cdata->padding;
- switch (ia->ri_memreg_strategy) {
- case RPCRDMA_FRMR:
- len += buf->rb_max_requests * RPCRDMA_MAX_SEGS *
- sizeof(struct rpcrdma_mw);
- break;
- case RPCRDMA_MTHCAFMR:
- /* TBD we are perhaps overallocating here */
- len += (buf->rb_max_requests + 1) * RPCRDMA_MAX_SEGS *
- sizeof(struct rpcrdma_mw);
- break;
- default:
- break;
- }
- /* allocate 1, 4 and 5 in one shot */
p = kzalloc(len, GFP_KERNEL);
if (p == NULL) {
dprintk("RPC: %s: req_t/rep_t/pad kzalloc(%zd) failed\n",
@@ -1075,53 +1140,16 @@ rpcrdma_buffer_create(struct rpcrdma_buffer *buf, struct rpcrdma_ep *ep,
INIT_LIST_HEAD(&buf->rb_mws);
INIT_LIST_HEAD(&buf->rb_all);
- r = (struct rpcrdma_mw *)p;
switch (ia->ri_memreg_strategy) {
case RPCRDMA_FRMR:
- for (i = buf->rb_max_requests * RPCRDMA_MAX_SEGS; i; i--) {
- r->r.frmr.fr_mr = ib_alloc_fast_reg_mr(ia->ri_pd,
- ia->ri_max_frmr_depth);
- if (IS_ERR(r->r.frmr.fr_mr)) {
- rc = PTR_ERR(r->r.frmr.fr_mr);
- dprintk("RPC: %s: ib_alloc_fast_reg_mr"
- " failed %i\n", __func__, rc);
- goto out;
- }
- r->r.frmr.fr_pgl = ib_alloc_fast_reg_page_list(
- ia->ri_id->device,
- ia->ri_max_frmr_depth);
- if (IS_ERR(r->r.frmr.fr_pgl)) {
- rc = PTR_ERR(r->r.frmr.fr_pgl);
- dprintk("RPC: %s: "
- "ib_alloc_fast_reg_page_list "
- "failed %i\n", __func__, rc);
-
- ib_dereg_mr(r->r.frmr.fr_mr);
- goto out;
- }
- list_add(&r->mw_all, &buf->rb_all);
- list_add(&r->mw_list, &buf->rb_mws);
- ++r;
- }
+ rc = rpcrdma_init_frmrs(ia, buf);
+ if (rc)
+ goto out;
break;
case RPCRDMA_MTHCAFMR:
- /* TBD we are perhaps overallocating here */
- for (i = (buf->rb_max_requests+1) * RPCRDMA_MAX_SEGS; i; i--) {
- static struct ib_fmr_attr fa =
- { RPCRDMA_MAX_DATA_SEGS, 1, PAGE_SHIFT };
- r->r.fmr = ib_alloc_fmr(ia->ri_pd,
- IB_ACCESS_REMOTE_WRITE | IB_ACCESS_REMOTE_READ,
- &fa);
- if (IS_ERR(r->r.fmr)) {
- rc = PTR_ERR(r->r.fmr);
- dprintk("RPC: %s: ib_alloc_fmr"
- " failed %i\n", __func__, rc);
- goto out;
- }
- list_add(&r->mw_all, &buf->rb_all);
- list_add(&r->mw_list, &buf->rb_mws);
- ++r;
- }
+ rc = rpcrdma_init_fmrs(ia, buf);
+ if (rc)
+ goto out;
break;
default:
break;
@@ -1189,24 +1217,57 @@ out:
return rc;
}
-/*
- * Unregister and destroy buffer memory. Need to deal with
- * partial initialization, so it's callable from failed create.
- * Must be called before destroying endpoint, as registrations
- * reference it.
- */
+static void
+rpcrdma_destroy_fmrs(struct rpcrdma_buffer *buf)
+{
+ struct rpcrdma_mw *r;
+ int rc;
+
+ while (!list_empty(&buf->rb_all)) {
+ r = list_entry(buf->rb_all.next, struct rpcrdma_mw, mw_all);
+ list_del(&r->mw_all);
+ list_del(&r->mw_list);
+
+ rc = ib_dealloc_fmr(r->r.fmr);
+ if (rc)
+ dprintk("RPC: %s: ib_dealloc_fmr failed %i\n",
+ __func__, rc);
+
+ kfree(r);
+ }
+}
+
+static void
+rpcrdma_destroy_frmrs(struct rpcrdma_buffer *buf)
+{
+ struct rpcrdma_mw *r;
+ int rc;
+
+ while (!list_empty(&buf->rb_all)) {
+ r = list_entry(buf->rb_all.next, struct rpcrdma_mw, mw_all);
+ list_del(&r->mw_all);
+ list_del(&r->mw_list);
+
+ rc = ib_dereg_mr(r->r.frmr.fr_mr);
+ if (rc)
+ dprintk("RPC: %s: ib_dereg_mr failed %i\n",
+ __func__, rc);
+ ib_free_fast_reg_page_list(r->r.frmr.fr_pgl);
+
+ kfree(r);
+ }
+}
+
void
rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf)
{
- int rc, i;
struct rpcrdma_ia *ia = rdmab_to_ia(buf);
- struct rpcrdma_mw *r;
+ int i;
/* clean up in reverse order from create
* 1. recv mr memory (mr free, then kfree)
* 2. send mr memory (mr free, then kfree)
- * 3. padding (if any) [moved to rpcrdma_ep_destroy]
- * 4. arrays
+ * 3. MWs
*/
dprintk("RPC: %s: entering\n", __func__);
@@ -1225,32 +1286,15 @@ rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf)
}
}
- while (!list_empty(&buf->rb_mws)) {
- r = list_entry(buf->rb_mws.next,
- struct rpcrdma_mw, mw_list);
- list_del(&r->mw_all);
- list_del(&r->mw_list);
- switch (ia->ri_memreg_strategy) {
- case RPCRDMA_FRMR:
- rc = ib_dereg_mr(r->r.frmr.fr_mr);
- if (rc)
- dprintk("RPC: %s:"
- " ib_dereg_mr"
- " failed %i\n",
- __func__, rc);
- ib_free_fast_reg_page_list(r->r.frmr.fr_pgl);
- break;
- case RPCRDMA_MTHCAFMR:
- rc = ib_dealloc_fmr(r->r.fmr);
- if (rc)
- dprintk("RPC: %s:"
- " ib_dealloc_fmr"
- " failed %i\n",
- __func__, rc);
- break;
- default:
- break;
- }
+ switch (ia->ri_memreg_strategy) {
+ case RPCRDMA_FRMR:
+ rpcrdma_destroy_frmrs(buf);
+ break;
+ case RPCRDMA_MTHCAFMR:
+ rpcrdma_destroy_fmrs(buf);
+ break;
+ default:
+ break;
}
kfree(buf->rb_pool);
next prev parent reply other threads:[~2014-07-22 14:53 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-07-22 14:50 [PATCH v4 00/21] NFS/RDMA client patches for 3.17 Chuck Lever
2014-07-22 14:51 ` [PATCH v4 01/21] xprtrdma: Fix panic in rpcrdma_register_frmr_external() Chuck Lever
2014-07-22 14:51 ` [PATCH v4 02/21] xprtrdma: Protect ia->ri_id when unmapping/invalidating MRs Chuck Lever
2014-07-22 14:51 ` [PATCH v4 03/21] xprtrdma: Limit data payload size for ALLPHYSICAL Chuck Lever
2014-07-22 14:51 ` [PATCH v4 04/21] xprtrdma: Update rkeys after transport reconnect Chuck Lever
2014-07-22 18:40 ` Anna Schumaker
2014-07-24 20:03 ` Chuck Lever
2014-07-22 14:51 ` [PATCH v4 05/21] xprtrdma: On disconnect, don't ignore pending CQEs Chuck Lever
2014-07-22 14:51 ` [PATCH v4 06/21] xprtrdma: Don't invalidate FRMRs if registration fails Chuck Lever
2014-07-22 14:51 ` [PATCH v4 07/21] xprtrdma: Unclutter struct rpcrdma_mr_seg Chuck Lever
2014-07-22 14:52 ` [PATCH v4 08/21] xprtrdma: Back off rkey when FAST_REG_MR fails Chuck Lever
2014-07-22 20:03 ` Anna Schumaker
2014-07-22 21:18 ` Chuck Lever
2014-07-23 13:44 ` Anna Schumaker
2014-07-24 13:24 ` Anna Schumaker
2014-07-30 19:11 ` Anna Schumaker
2014-07-22 14:52 ` [PATCH v4 09/21] xprtrdma: Chain together all MWs in same buffer pool Chuck Lever
2014-07-22 14:52 ` [PATCH v4 10/21] xprtrdma: Properly handle exhaustion of the rb_mws list Chuck Lever
2014-07-22 14:52 ` [PATCH v4 11/21] xprtrdma: Reset FRMRs when FAST_REG_MR is flushed by a disconnect Chuck Lever
2014-07-22 14:52 ` [PATCH v4 12/21] xprtrdma: Reset FRMRs after a flushed LOCAL_INV Work Request Chuck Lever
2014-07-22 14:52 ` [PATCH v4 13/21] xprtrdma: Don't post a LOCAL_INV in rpcrdma_register_frmr_external() Chuck Lever
2014-07-22 14:53 ` [PATCH v4 14/21] xprtrdma: Disable completions for FAST_REG_MR Work Requests Chuck Lever
2014-07-22 14:53 ` [PATCH v4 15/21] xprtrdma: Disable completions for LOCAL_INV " Chuck Lever
2014-07-22 14:53 ` [PATCH v4 16/21] xprtrdma: Rename frmr_wr Chuck Lever
2014-07-22 14:53 ` Chuck Lever [this message]
2014-07-22 14:53 ` [PATCH v4 18/21] xprtrdma: Schedule reply tasklet once per upcall Chuck Lever
2014-07-22 14:53 ` [PATCH v4 19/21] xprtrdma: Make rpcrdma_ep_disconnect() return void Chuck Lever
2014-07-22 14:53 ` [PATCH v4 20/21] xprtrdma: Remove RPCRDMA_PERSISTENT_REGISTRATION macro Chuck Lever
2014-07-22 14:54 ` [PATCH v4 21/21] xprtrdma: Handle additional connection events Chuck Lever
2014-07-23 18:53 ` [PATCH v4 00/21] NFS/RDMA client patches for 3.17 Anna Schumaker
2014-07-23 21:23 ` Chuck Lever
2014-07-24 13:19 ` Anna Schumaker
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20140722145327.6010.85935.stgit@manet.1015granger.net \
--to=chuck.lever@oracle.com \
--cc=Anna.Schumaker@netapp.com \
--cc=linux-nfs@vger.kernel.org \
--cc=linux-rdma@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).