summaryrefslogtreecommitdiffstats
path: root/fs/nfs/direct.c
diff options
context:
space:
mode:
authorTrond Myklebust <trond.myklebust@primarydata.com>2015-02-03 22:01:27 +0100
committerTrond Myklebust <trond.myklebust@primarydata.com>2015-02-03 22:01:27 +0100
commite2c63e091e29786a34ecf42c169e627a3d1d96d7 (patch)
tree4e96723b969bd6b3d39613de601f18da678cca71 /fs/nfs/direct.c
parentMerge tag 'nfs-rdma-for-3.20' of git://git.linux-nfs.org/projects/anna/nfs-rdma (diff)
parentpnfs: lookup new lseg at lseg boundary (diff)
downloadlinux-e2c63e091e29786a34ecf42c169e627a3d1d96d7.tar.xz
linux-e2c63e091e29786a34ecf42c169e627a3d1d96d7.zip
Merge branch 'flexfiles'
* flexfiles: (53 commits) pnfs: lookup new lseg at lseg boundary nfs41: .init_read and .init_write can be called with valid pg_lseg pnfs: Update documentation on the Layout Drivers pnfs/flexfiles: Add the FlexFile Layout Driver nfs: count DIO good bytes correctly with mirroring nfs41: wait for LAYOUTRETURN before retrying LAYOUTGET nfs: add a helper to set NFS_ODIRECT_RESCHED_WRITES to direct writes nfs41: add NFS_LAYOUT_RETRY_LAYOUTGET to layout header flags nfs/flexfiles: send layoutreturn before freeing lseg nfs41: introduce NFS_LAYOUT_RETURN_BEFORE_CLOSE nfs41: allow async version layoutreturn nfs41: add range to layoutreturn args pnfs: allow LD to ask to resend read through pnfs nfs: add nfs_pgio_current_mirror helper nfs: only reset desc->pg_mirror_idx when mirroring is supported nfs41: add a debug warning if we destroy an unempty layout pnfs: fail comparison when bucket verifier not set nfs: mirroring support for direct io nfs: add mirroring support to pgio layer pnfs: pass ds_commit_idx through the commit path ... Conflicts: fs/nfs/pnfs.c fs/nfs/pnfs.h
Diffstat (limited to 'fs/nfs/direct.c')
-rw-r--r--fs/nfs/direct.c112
1 files changed, 95 insertions, 17 deletions
diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c
index 294692ff83b1..7077521acdf4 100644
--- a/fs/nfs/direct.c
+++ b/fs/nfs/direct.c
@@ -66,6 +66,10 @@ static struct kmem_cache *nfs_direct_cachep;
/*
* This represents a set of asynchronous requests that we're waiting on
*/
+struct nfs_direct_mirror {
+ ssize_t count;
+};
+
struct nfs_direct_req {
struct kref kref; /* release manager */
@@ -78,8 +82,13 @@ struct nfs_direct_req {
/* completion state */
atomic_t io_count; /* i/os we're waiting for */
spinlock_t lock; /* protect completion state */
+
+ struct nfs_direct_mirror mirrors[NFS_PAGEIO_DESCRIPTOR_MIRROR_MAX];
+ int mirror_count;
+
ssize_t count, /* bytes actually processed */
bytes_left, /* bytes left to be sent */
+ io_start, /* start of IO */
error; /* any reported error */
struct completion completion; /* wait for i/o completion */
@@ -108,26 +117,56 @@ static inline int put_dreq(struct nfs_direct_req *dreq)
return atomic_dec_and_test(&dreq->io_count);
}
+void nfs_direct_set_resched_writes(struct nfs_direct_req *dreq)
+{
+ dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
+}
+EXPORT_SYMBOL_GPL(nfs_direct_set_resched_writes);
+
+static void
+nfs_direct_good_bytes(struct nfs_direct_req *dreq, struct nfs_pgio_header *hdr)
+{
+ int i;
+ ssize_t count;
+
+ WARN_ON_ONCE(hdr->pgio_mirror_idx >= dreq->mirror_count);
+
+ count = dreq->mirrors[hdr->pgio_mirror_idx].count;
+ if (count + dreq->io_start < hdr->io_start + hdr->good_bytes) {
+ count = hdr->io_start + hdr->good_bytes - dreq->io_start;
+ dreq->mirrors[hdr->pgio_mirror_idx].count = count;
+ }
+
+ /* update the dreq->count by finding the minimum agreed count from all
+ * mirrors */
+ count = dreq->mirrors[0].count;
+
+ for (i = 1; i < dreq->mirror_count; i++)
+ count = min(count, dreq->mirrors[i].count);
+
+ dreq->count = count;
+}
+
/*
* nfs_direct_select_verf - select the right verifier
* @dreq - direct request possibly spanning multiple servers
* @ds_clp - nfs_client of data server or NULL if MDS / non-pnfs
- * @ds_idx - index of data server in data server list, only valid if ds_clp set
+ * @commit_idx - commit bucket index for the DS
*
* returns the correct verifier to use given the role of the server
*/
static struct nfs_writeverf *
nfs_direct_select_verf(struct nfs_direct_req *dreq,
struct nfs_client *ds_clp,
- int ds_idx)
+ int commit_idx)
{
struct nfs_writeverf *verfp = &dreq->verf;
#ifdef CONFIG_NFS_V4_1
if (ds_clp) {
/* pNFS is in use, use the DS verf */
- if (ds_idx >= 0 && ds_idx < dreq->ds_cinfo.nbuckets)
- verfp = &dreq->ds_cinfo.buckets[ds_idx].direct_verf;
+ if (commit_idx >= 0 && commit_idx < dreq->ds_cinfo.nbuckets)
+ verfp = &dreq->ds_cinfo.buckets[commit_idx].direct_verf;
else
WARN_ON_ONCE(1);
}
@@ -148,8 +187,7 @@ static void nfs_direct_set_hdr_verf(struct nfs_direct_req *dreq,
{
struct nfs_writeverf *verfp;
- verfp = nfs_direct_select_verf(dreq, hdr->ds_clp,
- hdr->ds_idx);
+ verfp = nfs_direct_select_verf(dreq, hdr->ds_clp, hdr->ds_commit_idx);
WARN_ON_ONCE(verfp->committed >= 0);
memcpy(verfp, &hdr->verf, sizeof(struct nfs_writeverf));
WARN_ON_ONCE(verfp->committed < 0);
@@ -169,8 +207,7 @@ static int nfs_direct_set_or_cmp_hdr_verf(struct nfs_direct_req *dreq,
{
struct nfs_writeverf *verfp;
- verfp = nfs_direct_select_verf(dreq, hdr->ds_clp,
- hdr->ds_idx);
+ verfp = nfs_direct_select_verf(dreq, hdr->ds_clp, hdr->ds_commit_idx);
if (verfp->committed < 0) {
nfs_direct_set_hdr_verf(dreq, hdr);
return 0;
@@ -193,7 +230,11 @@ static int nfs_direct_cmp_commit_data_verf(struct nfs_direct_req *dreq,
verfp = nfs_direct_select_verf(dreq, data->ds_clp,
data->ds_commit_index);
- WARN_ON_ONCE(verfp->committed < 0);
+
+ /* verifier not set so always fail */
+ if (verfp->committed < 0)
+ return 1;
+
return memcmp(verfp, &data->verf, sizeof(struct nfs_writeverf));
}
@@ -249,6 +290,18 @@ void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo,
cinfo->completion_ops = &nfs_direct_commit_completion_ops;
}
+static inline void nfs_direct_setup_mirroring(struct nfs_direct_req *dreq,
+ struct nfs_pageio_descriptor *pgio,
+ struct nfs_page *req)
+{
+ int mirror_count = 1;
+
+ if (pgio->pg_ops->pg_get_mirror_count)
+ mirror_count = pgio->pg_ops->pg_get_mirror_count(pgio, req);
+
+ dreq->mirror_count = mirror_count;
+}
+
static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
{
struct nfs_direct_req *dreq;
@@ -263,6 +316,7 @@ static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
INIT_LIST_HEAD(&dreq->mds_cinfo.list);
dreq->verf.committed = NFS_INVALID_STABLE_HOW; /* not set yet */
INIT_WORK(&dreq->work, nfs_direct_write_schedule_work);
+ dreq->mirror_count = 1;
spin_lock_init(&dreq->lock);
return dreq;
@@ -369,7 +423,8 @@ static void nfs_direct_read_completion(struct nfs_pgio_header *hdr)
if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && (hdr->good_bytes == 0))
dreq->error = hdr->error;
else
- dreq->count += hdr->good_bytes;
+ nfs_direct_good_bytes(dreq, hdr);
+
spin_unlock(&dreq->lock);
while (!list_empty(&hdr->pages)) {
@@ -547,6 +602,7 @@ ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter,
dreq->inode = inode;
dreq->bytes_left = count;
+ dreq->io_start = pos;
dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
l_ctx = nfs_get_lock_context(dreq->ctx);
if (IS_ERR(l_ctx)) {
@@ -579,6 +635,20 @@ out:
return result;
}
+static void
+nfs_direct_write_scan_commit_list(struct inode *inode,
+ struct list_head *list,
+ struct nfs_commit_info *cinfo)
+{
+ spin_lock(cinfo->lock);
+#ifdef CONFIG_NFS_V4_1
+ if (cinfo->ds != NULL && cinfo->ds->nwritten != 0)
+ NFS_SERVER(inode)->pnfs_curr_ld->recover_commit_reqs(list, cinfo);
+#endif
+ nfs_scan_commit_list(&cinfo->mds->list, list, cinfo, 0);
+ spin_unlock(cinfo->lock);
+}
+
static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
{
struct nfs_pageio_descriptor desc;
@@ -586,20 +656,23 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
LIST_HEAD(reqs);
struct nfs_commit_info cinfo;
LIST_HEAD(failed);
+ int i;
nfs_init_cinfo_from_dreq(&cinfo, dreq);
- pnfs_recover_commit_reqs(dreq->inode, &reqs, &cinfo);
- spin_lock(cinfo.lock);
- nfs_scan_commit_list(&cinfo.mds->list, &reqs, &cinfo, 0);
- spin_unlock(cinfo.lock);
+ nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo);
dreq->count = 0;
+ for (i = 0; i < dreq->mirror_count; i++)
+ dreq->mirrors[i].count = 0;
get_dreq(dreq);
nfs_pageio_init_write(&desc, dreq->inode, FLUSH_STABLE, false,
&nfs_direct_write_completion_ops);
desc.pg_dreq = dreq;
+ req = nfs_list_entry(reqs.next);
+ nfs_direct_setup_mirroring(dreq, &desc, req);
+
list_for_each_entry_safe(req, tmp, &reqs, wb_list) {
if (!nfs_pageio_add_request(&desc, req)) {
nfs_list_remove_request(req);
@@ -646,7 +719,7 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data)
nfs_list_remove_request(req);
if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) {
/* Note the rewrite will go through mds */
- nfs_mark_request_commit(req, NULL, &cinfo);
+ nfs_mark_request_commit(req, NULL, &cinfo, 0);
} else
nfs_release_request(req);
nfs_unlock_and_release_request(req);
@@ -721,7 +794,7 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr)
dreq->error = hdr->error;
}
if (dreq->error == 0) {
- dreq->count += hdr->good_bytes;
+ nfs_direct_good_bytes(dreq, hdr);
if (nfs_write_need_commit(hdr)) {
if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES)
request_commit = true;
@@ -745,7 +818,8 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr)
nfs_list_remove_request(req);
if (request_commit) {
kref_get(&req->wb_kref);
- nfs_mark_request_commit(req, hdr->lseg, &cinfo);
+ nfs_mark_request_commit(req, hdr->lseg, &cinfo,
+ hdr->ds_commit_idx);
}
nfs_unlock_and_release_request(req);
}
@@ -826,6 +900,9 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
result = PTR_ERR(req);
break;
}
+
+ nfs_direct_setup_mirroring(dreq, &desc, req);
+
nfs_lock_request(req);
req->wb_index = pos >> PAGE_SHIFT;
req->wb_offset = pos & ~PAGE_MASK;
@@ -934,6 +1011,7 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter,
dreq->inode = inode;
dreq->bytes_left = count;
+ dreq->io_start = pos;
dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
l_ctx = nfs_get_lock_context(dreq->ctx);
if (IS_ERR(l_ctx)) {