* [Qemu-devel] [PATCH 1/6] hw/9pfs: Add reference counting for fid
2011-08-25 15:03 [Qemu-devel] [PATCH 0/6] hw/9pfs: Implement file descriptor reclaim in VirtFS server Aneesh Kumar K.V
@ 2011-08-25 15:03 ` Aneesh Kumar K.V
2011-08-25 15:03 ` [Qemu-devel] [PATCH 2/6] hw/9pfs: Add file descriptor reclaim support Aneesh Kumar K.V
` (4 subsequent siblings)
5 siblings, 0 replies; 7+ messages in thread
From: Aneesh Kumar K.V @ 2011-08-25 15:03 UTC (permalink / raw)
To: qemu-devel; +Cc: aliguori, Aneesh Kumar K.V
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
hw/9pfs/virtio-9p.c | 273 ++++++++++++++++++++++++++++++++++-----------------
hw/9pfs/virtio-9p.h | 2 +
2 files changed, 184 insertions(+), 91 deletions(-)
diff --git a/hw/9pfs/virtio-9p.c b/hw/9pfs/virtio-9p.c
index ad70768..555f456 100644
--- a/hw/9pfs/virtio-9p.c
+++ b/hw/9pfs/virtio-9p.c
@@ -232,16 +232,17 @@ static size_t v9fs_string_size(V9fsString *str)
return str->size;
}
-static V9fsFidState *lookup_fid(V9fsState *s, int32_t fid)
+static V9fsFidState *get_fid(V9fsState *s, int32_t fid)
{
V9fsFidState *f;
for (f = s->fid_list; f; f = f->next) {
+ BUG_ON(f->clunked);
if (f->fid == fid) {
+ f->ref++;
return f;
}
}
-
return NULL;
}
@@ -249,16 +250,17 @@ static V9fsFidState *alloc_fid(V9fsState *s, int32_t fid)
{
V9fsFidState *f;
- f = lookup_fid(s, fid);
- if (f) {
- return NULL;
+ for (f = s->fid_list; f; f = f->next) {
+ /* If fid is already there return NULL */
+ BUG_ON(f->clunked);
+ if (f->fid == fid) {
+ return NULL;
+ }
}
-
f = g_malloc0(sizeof(V9fsFidState));
-
f->fid = fid;
f->fid_type = P9_FID_NONE;
-
+ f->ref = 1;
f->next = s->fid_list;
s->fid_list = f;
@@ -299,9 +301,33 @@ free_value:
return retval;
}
-static int free_fid(V9fsState *s, int32_t fid)
+static int free_fid(V9fsState *s, V9fsFidState *fidp)
{
int retval = 0;
+
+ if (fidp->fid_type == P9_FID_FILE) {
+ retval = v9fs_co_close(s, fidp);
+ } else if (fidp->fid_type == P9_FID_DIR) {
+ retval = v9fs_co_closedir(s, fidp);
+ } else if (fidp->fid_type == P9_FID_XATTR) {
+ retval = v9fs_xattr_fid_clunk(s, fidp);
+ }
+ v9fs_string_free(&fidp->path);
+ g_free(fidp);
+ return retval;
+}
+
+static void put_fid(V9fsState *s, V9fsFidState *fidp)
+{
+ BUG_ON(!fidp->ref);
+ fidp->ref--;
+ if (!fidp->ref && fidp->clunked) {
+ free_fid(s, fidp);
+ }
+}
+
+static int clunk_fid(V9fsState *s, int32_t fid)
+{
V9fsFidState **fidpp, *fidp;
for (fidpp = &s->fid_list; *fidpp; fidpp = &(*fidpp)->next) {
@@ -313,20 +339,10 @@ static int free_fid(V9fsState *s, int32_t fid)
if (*fidpp == NULL) {
return -ENOENT;
}
-
fidp = *fidpp;
*fidpp = fidp->next;
-
- if (fidp->fid_type == P9_FID_FILE) {
- retval = v9fs_co_close(s, fidp);
- } else if (fidp->fid_type == P9_FID_DIR) {
- retval = v9fs_co_closedir(s, fidp);
- } else if (fidp->fid_type == P9_FID_XATTR) {
- retval = v9fs_xattr_fid_clunk(s, fidp);
- }
- v9fs_string_free(&fidp->path);
- g_free(fidp);
- return retval;
+ fidp->clunked = 1;
+ return 0;
}
#define P9_QID_TYPE_DIR 0x80
@@ -1014,19 +1030,21 @@ static void v9fs_attach(void *opaque)
fidp = alloc_fid(s, fid);
if (fidp == NULL) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
fidp->uid = n_uname;
v9fs_string_sprintf(&fidp->path, "%s", "/");
err = fid_to_qid(s, fidp, &qid);
if (err < 0) {
err = -EINVAL;
- free_fid(s, fid);
+ clunk_fid(s, fid);
goto out;
}
offset += pdu_marshal(pdu, offset, "Q", &qid);
err = offset;
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
v9fs_string_free(&uname);
v9fs_string_free(&aname);
@@ -1044,10 +1062,11 @@ static void v9fs_stat(void *opaque)
V9fsState *s = pdu->s;
pdu_unmarshal(pdu, offset, "d", &fid);
- fidp = lookup_fid(s, fid);
+
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
err = v9fs_co_lstat(s, &fidp->path, &stbuf);
if (err < 0) {
@@ -1061,6 +1080,8 @@ static void v9fs_stat(void *opaque)
err = offset;
v9fs_stat_free(&v9stat);
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
}
@@ -1078,10 +1099,10 @@ static void v9fs_getattr(void *opaque)
pdu_unmarshal(pdu, offset, "dq", &fid, &request_mask);
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
retval = -ENOENT;
- goto out;
+ goto out_nofid;
}
/*
* Currently we only support BASIC fields in stat, so there is no
@@ -1095,6 +1116,8 @@ static void v9fs_getattr(void *opaque)
retval = offset;
retval += pdu_marshal(pdu, offset, "A", &v9stat_dotl);
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, retval);
}
@@ -1122,10 +1145,10 @@ static void v9fs_setattr(void *opaque)
pdu_unmarshal(pdu, offset, "dI", &fid, &v9iattr);
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
if (v9iattr.valid & ATTR_MODE) {
err = v9fs_co_chmod(s, &fidp->path, v9iattr.mode);
@@ -1187,6 +1210,8 @@ static void v9fs_setattr(void *opaque)
}
err = offset;
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
}
@@ -1213,7 +1238,7 @@ static void v9fs_walk(void *opaque)
int32_t fid, newfid;
V9fsString *wnames = NULL;
V9fsFidState *fidp;
- V9fsFidState *newfidp;
+ V9fsFidState *newfidp = NULL;;
V9fsPDU *pdu = opaque;
V9fsState *s = pdu->s;
@@ -1229,12 +1254,12 @@ static void v9fs_walk(void *opaque)
} else if (nwnames > P9_MAXWELEM) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
if (fid == newfid) {
BUG_ON(fidp->fid_type != P9_FID_NONE);
@@ -1267,7 +1292,7 @@ static void v9fs_walk(void *opaque)
v9fs_string_copy(&newfidp->path, &path);
err = v9fs_co_lstat(s, &newfidp->path, &stbuf);
if (err < 0) {
- free_fid(s, newfidp->fid);
+ clunk_fid(s, newfidp->fid);
v9fs_string_free(&path);
goto out;
}
@@ -1277,6 +1302,11 @@ static void v9fs_walk(void *opaque)
}
err = v9fs_walk_marshal(pdu, nwnames, qids);
out:
+ put_fid(s, fidp);
+ if (newfidp) {
+ put_fid(s, newfidp);
+ }
+out_nofid:
complete_pdu(s, pdu, err);
if (nwnames && nwnames <= P9_MAXWELEM) {
for (name_idx = 0; name_idx < nwnames; name_idx++) {
@@ -1325,10 +1355,10 @@ static void v9fs_open(void *opaque)
} else {
pdu_unmarshal(pdu, offset, "db", &fid, &mode);
}
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
BUG_ON(fidp->fid_type != P9_FID_NONE);
@@ -1364,6 +1394,8 @@ static void v9fs_open(void *opaque)
err = offset;
}
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
}
@@ -1385,10 +1417,10 @@ static void v9fs_lcreate(void *opaque)
pdu_unmarshal(pdu, offset, "dsddd", &dfid, &name, &flags,
&mode, &gid);
- fidp = lookup_fid(pdu->s, dfid);
+ fidp = get_fid(pdu->s, dfid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
v9fs_string_sprintf(&fullname, "%s/%s", fidp->path.data, name.data);
@@ -1415,6 +1447,8 @@ static void v9fs_lcreate(void *opaque)
offset += pdu_marshal(pdu, offset, "Qd", &qid, iounit);
err = offset;
out:
+ put_fid(pdu->s, fidp);
+out_nofid:
complete_pdu(pdu->s, pdu, err);
v9fs_string_free(&name);
v9fs_string_free(&fullname);
@@ -1431,16 +1465,17 @@ static void v9fs_fsync(void *opaque)
V9fsState *s = pdu->s;
pdu_unmarshal(pdu, offset, "dd", &fid, &datasync);
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
err = v9fs_co_fsync(s, fidp, datasync);
if (!err) {
err = offset;
}
-out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
}
@@ -1449,16 +1484,25 @@ static void v9fs_clunk(void *opaque)
int err;
int32_t fid;
size_t offset = 7;
+ V9fsFidState *fidp;
V9fsPDU *pdu = opaque;
V9fsState *s = pdu->s;
pdu_unmarshal(pdu, offset, "d", &fid);
- err = free_fid(s, fid);
+
+ fidp = get_fid(s, fid);
+ if (fidp == NULL) {
+ err = -ENOENT;
+ goto out_nofid;
+ }
+ err = clunk_fid(s, fidp->fid);
if (err < 0) {
goto out;
}
err = offset;
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
}
@@ -1557,10 +1601,11 @@ static void v9fs_read(void *opaque)
V9fsState *s = pdu->s;
pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &max_count);
- fidp = lookup_fid(s, fid);
+
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
if (fidp->fid_type == P9_FID_DIR) {
@@ -1612,6 +1657,8 @@ static void v9fs_read(void *opaque)
err = -EINVAL;
}
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
}
@@ -1696,8 +1743,12 @@ static void v9fs_readdir(void *opaque)
pdu_unmarshal(pdu, offset, "dqd", &fid, &initial_offset, &max_count);
- fidp = lookup_fid(s, fid);
- if (fidp == NULL || !fidp->fs.dir) {
+ fidp = get_fid(s, fid);
+ if (fidp == NULL) {
+ retval = -EINVAL;
+ goto out_nofid;
+ }
+ if (!fidp->fs.dir) {
retval = -EINVAL;
goto out;
}
@@ -1715,6 +1766,8 @@ static void v9fs_readdir(void *opaque)
retval += pdu_marshal(pdu, offset, "d", count);
retval += count;
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, retval);
}
@@ -1779,10 +1832,11 @@ static void v9fs_write(void *opaque)
V9fsState *s = pdu->s;
pdu_unmarshal(pdu, offset, "dqdv", &fid, &off, &count, sg, &cnt);
- fidp = lookup_fid(s, fid);
+
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
if (fidp->fid_type == P9_FID_FILE) {
if (fidp->fs.fd == -1) {
@@ -1822,6 +1876,8 @@ static void v9fs_write(void *opaque)
offset += pdu_marshal(pdu, offset, "d", total);
err = offset;
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
}
@@ -1846,10 +1902,10 @@ static void v9fs_create(void *opaque)
pdu_unmarshal(pdu, offset, "dsdbs", &fid, &name,
&perm, &mode, &extension);
- fidp = lookup_fid(pdu->s, fid);
+ fidp = get_fid(pdu->s, fid);
if (fidp == NULL) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
v9fs_string_sprintf(&fullname, "%s/%s", fidp->path.data, name.data);
@@ -1879,15 +1935,17 @@ static void v9fs_create(void *opaque)
}
} else if (perm & P9_STAT_MODE_LINK) {
int32_t nfid = atoi(extension.data);
- V9fsFidState *nfidp = lookup_fid(pdu->s, nfid);
+ V9fsFidState *nfidp = get_fid(pdu->s, nfid);
if (nfidp == NULL) {
err = -EINVAL;
goto out;
}
err = v9fs_co_link(pdu->s, &nfidp->path, &fullname);
if (err < 0) {
+ put_fid(pdu->s, nfidp);
goto out;
}
+ put_fid(pdu->s, nfidp);
} else if (perm & P9_STAT_MODE_DEVICE) {
char ctype;
uint32_t major, minor;
@@ -1950,6 +2008,8 @@ static void v9fs_create(void *opaque)
offset += pdu_marshal(pdu, offset, "Qd", &qid, iounit);
err = offset;
out:
+ put_fid(pdu->s, fidp);
+out_nofid:
complete_pdu(pdu->s, pdu, err);
v9fs_string_free(&name);
v9fs_string_free(&extension);
@@ -1973,10 +2033,10 @@ static void v9fs_symlink(void *opaque)
v9fs_string_init(&fullname);
pdu_unmarshal(pdu, offset, "dssd", &dfid, &name, &symname, &gid);
- dfidp = lookup_fid(pdu->s, dfid);
+ dfidp = get_fid(pdu->s, dfid);
if (dfidp == NULL) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
v9fs_string_sprintf(&fullname, "%s/%s", dfidp->path.data, name.data);
@@ -1992,6 +2052,8 @@ static void v9fs_symlink(void *opaque)
offset += pdu_marshal(pdu, offset, "Q", &qid);
err = offset;
out:
+ put_fid(pdu->s, dfidp);
+out_nofid:
complete_pdu(pdu->s, pdu, err);
v9fs_string_free(&name);
v9fs_string_free(&symname);
@@ -2021,13 +2083,13 @@ static void v9fs_link(void *opaque)
pdu_unmarshal(pdu, offset, "dds", &dfid, &oldfid, &name);
- dfidp = lookup_fid(s, dfid);
+ dfidp = get_fid(s, dfid);
if (dfidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
- oldfidp = lookup_fid(s, oldfid);
+ oldfidp = get_fid(s, oldfid);
if (oldfidp == NULL) {
err = -ENOENT;
goto out;
@@ -2041,6 +2103,8 @@ static void v9fs_link(void *opaque)
v9fs_string_free(&fullname);
out:
+ put_fid(s, dfidp);
+out_nofid:
v9fs_string_free(&name);
complete_pdu(s, pdu, err);
}
@@ -2055,10 +2119,10 @@ static void v9fs_remove(void *opaque)
pdu_unmarshal(pdu, offset, "d", &fid);
- fidp = lookup_fid(pdu->s, fid);
+ fidp = get_fid(pdu->s, fid);
if (fidp == NULL) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
err = v9fs_co_remove(pdu->s, &fidp->path);
if (!err) {
@@ -2066,8 +2130,9 @@ static void v9fs_remove(void *opaque)
}
/* For TREMOVE we need to clunk the fid even on failed remove */
- free_fid(pdu->s, fidp->fid);
-out:
+ clunk_fid(pdu->s, fidp->fid);
+ put_fid(pdu->s, fidp);
+out_nofid:
complete_pdu(pdu->s, pdu, err);
}
@@ -2076,14 +2141,14 @@ static int v9fs_complete_rename(V9fsState *s, V9fsFidState *fidp,
{
char *end;
int err = 0;
+ V9fsFidState *dirfidp = NULL;
char *old_name, *new_name;
if (newdirfid != -1) {
- V9fsFidState *dirfidp;
- dirfidp = lookup_fid(s, newdirfid);
+ dirfidp = get_fid(s, newdirfid);
if (dirfidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
BUG_ON(dirfidp->fid_type != P9_FID_NONE);
@@ -2136,6 +2201,10 @@ static int v9fs_complete_rename(V9fsState *s, V9fsFidState *fidp,
v9fs_string_copy(&fidp->path, name);
}
out:
+ if (dirfidp) {
+ put_fid(s, dirfidp);
+ }
+out_nofid:
return err;
}
@@ -2152,10 +2221,10 @@ static void v9fs_rename(void *opaque)
pdu_unmarshal(pdu, offset, "dds", &fid, &newdirfid, &name);
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
BUG_ON(fidp->fid_type != P9_FID_NONE);
@@ -2163,7 +2232,8 @@ static void v9fs_rename(void *opaque)
if (!err) {
err = offset;
}
-out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
v9fs_string_free(&name);
}
@@ -2181,10 +2251,11 @@ static void v9fs_wstat(void *opaque)
V9fsState *s = pdu->s;
pdu_unmarshal(pdu, offset, "dwS", &fid, &unused, &v9stat);
- fidp = lookup_fid(s, fid);
+
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
/* do we need to sync the file? */
if (donttouch_stat(&v9stat)) {
@@ -2250,6 +2321,8 @@ static void v9fs_wstat(void *opaque)
}
err = offset;
out:
+ put_fid(s, fidp);
+out_nofid:
v9fs_stat_free(&v9stat);
complete_pdu(s, pdu, err);
}
@@ -2310,10 +2383,10 @@ static void v9fs_statfs(void *opaque)
V9fsState *s = pdu->s;
pdu_unmarshal(pdu, offset, "d", &fid);
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
retval = -ENOENT;
- goto out;
+ goto out_nofid;
}
retval = v9fs_co_statfs(s, &fidp->path, &stbuf);
if (retval < 0) {
@@ -2322,6 +2395,8 @@ static void v9fs_statfs(void *opaque)
retval = offset;
retval += v9fs_fill_statfs(s, pdu, &stbuf);
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, retval);
return;
}
@@ -2347,10 +2422,10 @@ static void v9fs_mknod(void *opaque)
pdu_unmarshal(pdu, offset, "dsdddd", &fid, &name, &mode,
&major, &minor, &gid);
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
v9fs_string_sprintf(&fullname, "%s/%s", fidp->path.data, name.data);
err = v9fs_co_mknod(s, &fullname, fidp->uid, gid,
@@ -2366,6 +2441,8 @@ static void v9fs_mknod(void *opaque)
err = offset;
err += pdu_marshal(pdu, offset, "Q", &qid);
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
v9fs_string_free(&fullname);
v9fs_string_free(&name);
@@ -2399,12 +2476,12 @@ static void v9fs_lock(void *opaque)
/* We support only block flag now (that too ignored currently) */
if (flock->flags & ~P9_LOCK_FLAGS_BLOCK) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
err = v9fs_co_fstat(s, fidp->fs.fd, &stbuf);
if (err < 0) {
@@ -2412,6 +2489,8 @@ static void v9fs_lock(void *opaque)
}
status = P9_LOCK_SUCCESS;
out:
+ put_fid(s, fidp);
+out_nofid:
err = offset;
err += pdu_marshal(pdu, offset, "b", status);
complete_pdu(s, pdu, err);
@@ -2437,10 +2516,10 @@ static void v9fs_getlock(void *opaque)
&glock->start, &glock->length, &glock->proc_id,
&glock->client_id);
- fidp = lookup_fid(s, fid);
+ fidp = get_fid(s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
err = v9fs_co_fstat(s, fidp->fs.fd, &stbuf);
if (err < 0) {
@@ -2452,6 +2531,8 @@ static void v9fs_getlock(void *opaque)
&glock->client_id);
err = offset;
out:
+ put_fid(s, fidp);
+out_nofid:
complete_pdu(s, pdu, err);
g_free(glock);
}
@@ -2472,10 +2553,10 @@ static void v9fs_mkdir(void *opaque)
v9fs_string_init(&fullname);
pdu_unmarshal(pdu, offset, "dsdd", &fid, &name, &mode, &gid);
- fidp = lookup_fid(pdu->s, fid);
+ fidp = get_fid(pdu->s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
v9fs_string_sprintf(&fullname, "%s/%s", fidp->path.data, name.data);
err = v9fs_co_mkdir(pdu->s, fullname.data, mode, fidp->uid, gid);
@@ -2490,6 +2571,8 @@ static void v9fs_mkdir(void *opaque)
offset += pdu_marshal(pdu, offset, "Q", &qid);
err = offset;
out:
+ put_fid(pdu->s, fidp);
+out_nofid:
complete_pdu(pdu->s, pdu, err);
v9fs_string_free(&fullname);
v9fs_string_free(&name);
@@ -2503,15 +2586,15 @@ static void v9fs_xattrwalk(void *opaque)
size_t offset = 7;
int32_t fid, newfid;
V9fsFidState *file_fidp;
- V9fsFidState *xattr_fidp;
+ V9fsFidState *xattr_fidp = NULL;
V9fsPDU *pdu = opaque;
V9fsState *s = pdu->s;
pdu_unmarshal(pdu, offset, "dds", &fid, &newfid, &name);
- file_fidp = lookup_fid(s, fid);
+ file_fidp = get_fid(s, fid);
if (file_fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
xattr_fidp = alloc_fid(s, newfid);
if (xattr_fidp == NULL) {
@@ -2526,7 +2609,7 @@ static void v9fs_xattrwalk(void *opaque)
size = v9fs_co_llistxattr(s, &xattr_fidp->path, NULL, 0);
if (size < 0) {
err = size;
- free_fid(s, xattr_fidp->fid);
+ clunk_fid(s, xattr_fidp->fid);
goto out;
}
/*
@@ -2541,7 +2624,7 @@ static void v9fs_xattrwalk(void *opaque)
xattr_fidp->fs.xattr.value,
xattr_fidp->fs.xattr.len);
if (err < 0) {
- free_fid(s, xattr_fidp->fid);
+ clunk_fid(s, xattr_fidp->fid);
goto out;
}
}
@@ -2556,7 +2639,7 @@ static void v9fs_xattrwalk(void *opaque)
&name, NULL, 0);
if (size < 0) {
err = size;
- free_fid(s, xattr_fidp->fid);
+ clunk_fid(s, xattr_fidp->fid);
goto out;
}
/*
@@ -2571,7 +2654,7 @@ static void v9fs_xattrwalk(void *opaque)
&name, xattr_fidp->fs.xattr.value,
xattr_fidp->fs.xattr.len);
if (err < 0) {
- free_fid(s, xattr_fidp->fid);
+ clunk_fid(s, xattr_fidp->fid);
goto out;
}
}
@@ -2579,6 +2662,11 @@ static void v9fs_xattrwalk(void *opaque)
err = offset;
}
out:
+ put_fid(s, file_fidp);
+ if (xattr_fidp) {
+ put_fid(s, xattr_fidp);
+ }
+out_nofid:
complete_pdu(s, pdu, err);
v9fs_string_free(&name);
}
@@ -2599,10 +2687,10 @@ static void v9fs_xattrcreate(void *opaque)
pdu_unmarshal(pdu, offset, "dsqd",
&fid, &name, &size, &flags);
- file_fidp = lookup_fid(s, fid);
+ file_fidp = get_fid(s, fid);
if (file_fidp == NULL) {
err = -EINVAL;
- goto out;
+ goto out_nofid;
}
/* Make the file fid point to xattr */
xattr_fidp = file_fidp;
@@ -2618,7 +2706,8 @@ static void v9fs_xattrcreate(void *opaque)
xattr_fidp->fs.xattr.value = NULL;
}
err = offset;
-out:
+ put_fid(s, file_fidp);
+out_nofid:
complete_pdu(s, pdu, err);
v9fs_string_free(&name);
}
@@ -2633,10 +2722,10 @@ static void v9fs_readlink(void *opaque)
V9fsFidState *fidp;
pdu_unmarshal(pdu, offset, "d", &fid);
- fidp = lookup_fid(pdu->s, fid);
+ fidp = get_fid(pdu->s, fid);
if (fidp == NULL) {
err = -ENOENT;
- goto out;
+ goto out_nofid;
}
v9fs_string_init(&target);
@@ -2648,6 +2737,8 @@ static void v9fs_readlink(void *opaque)
err = offset;
v9fs_string_free(&target);
out:
+ put_fid(pdu->s, fidp);
+out_nofid:
complete_pdu(pdu->s, pdu, err);
}
diff --git a/hw/9pfs/virtio-9p.h b/hw/9pfs/virtio-9p.h
index 1d8c1b1..e16e5f4 100644
--- a/hw/9pfs/virtio-9p.h
+++ b/hw/9pfs/virtio-9p.h
@@ -203,6 +203,8 @@ struct V9fsFidState
V9fsXattr xattr;
} fs;
uid_t uid;
+ int ref;
+ int clunked;
V9fsFidState *next;
};
--
1.7.4.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* [Qemu-devel] [PATCH 2/6] hw/9pfs: Add file descriptor reclaim support
2011-08-25 15:03 [Qemu-devel] [PATCH 0/6] hw/9pfs: Implement file descriptor reclaim in VirtFS server Aneesh Kumar K.V
2011-08-25 15:03 ` [Qemu-devel] [PATCH 1/6] hw/9pfs: Add reference counting for fid Aneesh Kumar K.V
@ 2011-08-25 15:03 ` Aneesh Kumar K.V
2011-08-25 15:03 ` [Qemu-devel] [PATCH 3/6] hw/9pfs: init fid list properly Aneesh Kumar K.V
` (3 subsequent siblings)
5 siblings, 0 replies; 7+ messages in thread
From: Aneesh Kumar K.V @ 2011-08-25 15:03 UTC (permalink / raw)
To: qemu-devel; +Cc: aliguori, Aneesh Kumar K.V
[M. Mohan Kumar <mohan@in.ibm.com> removed some unused variables]
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
hw/9pfs/codir.c | 4 +-
hw/9pfs/cofile.c | 19 ++++-
hw/9pfs/virtio-9p-coth.h | 4 +-
hw/9pfs/virtio-9p-device.c | 1 +
hw/9pfs/virtio-9p.c | 187 +++++++++++++++++++++++++++++++++++++++++++-
hw/9pfs/virtio-9p.h | 22 +++++-
6 files changed, 223 insertions(+), 14 deletions(-)
diff --git a/hw/9pfs/codir.c b/hw/9pfs/codir.c
index 3a257b9..86adc87 100644
--- a/hw/9pfs/codir.c
+++ b/hw/9pfs/codir.c
@@ -100,12 +100,10 @@ int v9fs_co_opendir(V9fsState *s, V9fsFidState *fidp)
return err;
}
-int v9fs_co_closedir(V9fsState *s, V9fsFidState *fidp)
+int v9fs_co_closedir(V9fsState *s, DIR *dir)
{
int err;
- DIR *dir;
- dir = fidp->fs.dir;
v9fs_co_run_in_worker(
{
err = s->ops->closedir(&s->ctx, dir);
diff --git a/hw/9pfs/cofile.c b/hw/9pfs/cofile.c
index e388146..0caf1e3 100644
--- a/hw/9pfs/cofile.c
+++ b/hw/9pfs/cofile.c
@@ -58,6 +58,12 @@ int v9fs_co_open(V9fsState *s, V9fsFidState *fidp, int flags)
err = 0;
}
});
+ if (!err) {
+ total_open_fd++;
+ if (total_open_fd > open_fd_hw) {
+ v9fs_reclaim_fd(s);
+ }
+ }
return err;
}
@@ -79,15 +85,19 @@ int v9fs_co_open2(V9fsState *s, V9fsFidState *fidp, char *fullname, gid_t gid,
err = -errno;
}
});
+ if (!err) {
+ total_open_fd++;
+ if (total_open_fd > open_fd_hw) {
+ v9fs_reclaim_fd(s);
+ }
+ }
return err;
}
-int v9fs_co_close(V9fsState *s, V9fsFidState *fidp)
+int v9fs_co_close(V9fsState *s, int fd)
{
- int fd;
int err;
- fd = fidp->fs.fd;
v9fs_co_run_in_worker(
{
err = s->ops->close(&s->ctx, fd);
@@ -95,6 +105,9 @@ int v9fs_co_close(V9fsState *s, V9fsFidState *fidp)
err = -errno;
}
});
+ if (!err) {
+ total_open_fd--;
+ }
return err;
}
diff --git a/hw/9pfs/virtio-9p-coth.h b/hw/9pfs/virtio-9p-coth.h
index 48defb7..b7be9b5 100644
--- a/hw/9pfs/virtio-9p-coth.h
+++ b/hw/9pfs/virtio-9p-coth.h
@@ -83,8 +83,8 @@ extern int v9fs_co_open2(V9fsState *, V9fsFidState *, char *, gid_t, int, int);
extern int v9fs_co_lsetxattr(V9fsState *, V9fsString *, V9fsString *,
void *, size_t, int);
extern int v9fs_co_lremovexattr(V9fsState *, V9fsString *, V9fsString *);
-extern int v9fs_co_closedir(V9fsState *, V9fsFidState *);
-extern int v9fs_co_close(V9fsState *, V9fsFidState *);
+extern int v9fs_co_closedir(V9fsState *, DIR *);
+extern int v9fs_co_close(V9fsState *, int);
extern int v9fs_co_fsync(V9fsState *, V9fsFidState *, int);
extern int v9fs_co_symlink(V9fsState *, V9fsFidState *, const char *,
const char *, gid_t);
diff --git a/hw/9pfs/virtio-9p-device.c b/hw/9pfs/virtio-9p-device.c
index 9466002..44b9ace 100644
--- a/hw/9pfs/virtio-9p-device.c
+++ b/hw/9pfs/virtio-9p-device.c
@@ -171,6 +171,7 @@ static PCIDeviceInfo virtio_9p_info = {
static void virtio_9p_register_devices(void)
{
pci_qdev_register(&virtio_9p_info);
+ virtio_9p_set_fd_limit();
}
device_init(virtio_9p_register_devices)
diff --git a/hw/9pfs/virtio-9p.c b/hw/9pfs/virtio-9p.c
index 555f456..4ba2590 100644
--- a/hw/9pfs/virtio-9p.c
+++ b/hw/9pfs/virtio-9p.c
@@ -22,6 +22,9 @@
#include "virtio-9p-coth.h"
int debug_9p_pdu;
+int open_fd_hw;
+int total_open_fd;
+static int open_fd_rc;
enum {
Oread = 0x00,
@@ -234,12 +237,41 @@ static size_t v9fs_string_size(V9fsString *str)
static V9fsFidState *get_fid(V9fsState *s, int32_t fid)
{
+ int err;
V9fsFidState *f;
for (f = s->fid_list; f; f = f->next) {
BUG_ON(f->clunked);
if (f->fid == fid) {
+ /*
+ * Update the fid ref upfront so that
+ * we don't get reclaimed when we yield
+ * in open later.
+ */
f->ref++;
+
+ /*
+ * check whether we need to reopen the
+ * file. We might have closed the fd
+ * while trying to free up some file
+ * descriptors.
+ */
+ if (f->fid_type == P9_FID_FILE) {
+ if (f->fs.fd == -1) {
+ do {
+ err = v9fs_co_open(s, f, f->open_flags);
+ } while (err == -EINTR);
+ if (err < 0) {
+ f->ref--;
+ return NULL;
+ }
+ }
+ }
+ /*
+ * Mark the fid as referenced so that the LRU
+ * reclaim won't close the file descriptor
+ */
+ f->flags |= FID_REFERENCED;
return f;
}
}
@@ -261,6 +293,11 @@ static V9fsFidState *alloc_fid(V9fsState *s, int32_t fid)
f->fid = fid;
f->fid_type = P9_FID_NONE;
f->ref = 1;
+ /*
+ * Mark the fid as referenced so that the LRU
+ * reclaim won't close the file descriptor
+ */
+ f->flags |= FID_REFERENCED;
f->next = s->fid_list;
s->fid_list = f;
@@ -306,9 +343,12 @@ static int free_fid(V9fsState *s, V9fsFidState *fidp)
int retval = 0;
if (fidp->fid_type == P9_FID_FILE) {
- retval = v9fs_co_close(s, fidp);
+ /* If we reclaimed the fd no need to close */
+ if (fidp->fs.fd != -1) {
+ retval = v9fs_co_close(s, fidp->fs.fd);
+ }
} else if (fidp->fid_type == P9_FID_DIR) {
- retval = v9fs_co_closedir(s, fidp);
+ retval = v9fs_co_closedir(s, fidp->fs.dir);
} else if (fidp->fid_type == P9_FID_XATTR) {
retval = v9fs_xattr_fid_clunk(s, fidp);
}
@@ -321,6 +361,9 @@ static void put_fid(V9fsState *s, V9fsFidState *fidp)
{
BUG_ON(!fidp->ref);
fidp->ref--;
+ /*
+ * Don't free the fid if it is in reclaim list
+ */
if (!fidp->ref && fidp->clunked) {
free_fid(s, fidp);
}
@@ -345,6 +388,101 @@ static int clunk_fid(V9fsState *s, int32_t fid)
return 0;
}
+void v9fs_reclaim_fd(V9fsState *s)
+{
+ int reclaim_count = 0;
+ V9fsFidState *f, *reclaim_list = NULL;
+
+ for (f = s->fid_list; f; f = f->next) {
+ /*
+ * Unlink fids cannot be reclaimed. Check
+ * for them and skip them. Also skip fids
+ * currently being operated on.
+ */
+ if (f->ref || f->flags & FID_NON_RECLAIMABLE) {
+ continue;
+ }
+ /*
+ * if it is a recently referenced fid
+ * we leave the fid untouched and clear the
+ * reference bit. We come back to it later
+ * in the next iteration. (a simple LRU without
+ * moving list elements around)
+ */
+ if (f->flags & FID_REFERENCED) {
+ f->flags &= ~FID_REFERENCED;
+ continue;
+ }
+ /*
+ * Add fids to reclaim list.
+ */
+ if (f->fid_type == P9_FID_FILE) {
+ if (f->fs.fd != -1) {
+ /*
+ * Up the reference count so that
+ * a clunk request won't free this fid
+ */
+ f->ref++;
+ f->rclm_lst = reclaim_list;
+ reclaim_list = f;
+ f->fs_reclaim.fd = f->fs.fd;
+ f->fs.fd = -1;
+ reclaim_count++;
+ }
+ }
+ if (reclaim_count >= open_fd_rc) {
+ break;
+ }
+ }
+ /*
+ * Now close the fid in reclaim list. Free them if they
+ * are already clunked.
+ */
+ while (reclaim_list) {
+ f = reclaim_list;
+ reclaim_list = f->rclm_lst;
+ if (f->fid_type == P9_FID_FILE) {
+ v9fs_co_close(s, f->fs_reclaim.fd);
+ }
+ f->rclm_lst = NULL;
+ /*
+ * Now drop the fid reference, free it
+ * if clunked.
+ */
+ put_fid(s, f);
+ }
+}
+
+static int v9fs_mark_fids_unreclaim(V9fsState *s, V9fsString *str)
+{
+ int err;
+ V9fsFidState *fidp, head_fid;
+
+ head_fid.next = s->fid_list;
+ for (fidp = s->fid_list; fidp; fidp = fidp->next) {
+ if (!strcmp(fidp->path.data, str->data)) {
+ /* Mark the fid non reclaimable. */
+ fidp->flags |= FID_NON_RECLAIMABLE;
+ /* reopen the file if already closed */
+ if (fidp->fs.fd == -1) {
+ do {
+ err = v9fs_co_open(s, fidp, fidp->open_flags);
+ } while (err == -EINTR);
+ if (err < 0) {
+ return -1;
+ }
+ /*
+ * Go back to head of fid list because
+ * the list could have got updated when
+ * switched to the worker thread
+ */
+ fidp = &head_fid;
+ }
+ }
+ }
+ return 0;
+}
+
#define P9_QID_TYPE_DIR 0x80
#define P9_QID_TYPE_SYMLINK 0x02
@@ -1389,6 +1527,14 @@ static void v9fs_open(void *opaque)
goto out;
}
fidp->fid_type = P9_FID_FILE;
+ fidp->open_flags = flags;
+ if (flags & O_EXCL) {
+ /*
+ * We let the host file system do O_EXCL check
+ * We should not reclaim such fd
+ */
+ fidp->flags |= FID_NON_RECLAIMABLE;
+ }
iounit = get_iounit(s, &fidp->path);
offset += pdu_marshal(pdu, offset, "Qd", &qid, iounit);
err = offset;
@@ -1432,6 +1578,14 @@ static void v9fs_lcreate(void *opaque)
goto out;
}
fidp->fid_type = P9_FID_FILE;
+ fidp->open_flags = flags;
+ if (flags & O_EXCL) {
+ /*
+ * We let the host file system do O_EXCL check
+ * We should not reclaim such fd
+ */
+ fidp->flags |= FID_NON_RECLAIMABLE;
+ }
iounit = get_iounit(pdu->s, &fullname);
err = v9fs_co_lstat(pdu->s, &fullname, &stbuf);
@@ -1993,6 +2147,14 @@ static void v9fs_create(void *opaque)
goto out;
}
fidp->fid_type = P9_FID_FILE;
+ fidp->open_flags = omode_to_uflags(mode);
+ if (fidp->open_flags & O_EXCL) {
+ /*
+ * We let the host file system do O_EXCL check
+ * We should not reclaim such fd
+ */
+ fidp->flags |= FID_NON_RECLAIMABLE;
+ }
}
err = v9fs_co_lstat(pdu->s, &fullname, &stbuf);
if (err < 0) {
@@ -2124,11 +2286,19 @@ static void v9fs_remove(void *opaque)
err = -EINVAL;
goto out_nofid;
}
+ /*
+ * IF the file is unlinked, we cannot reopen
+ * the file later. So don't reclaim fd
+ */
+ err = v9fs_mark_fids_unreclaim(pdu->s, &fidp->path);
+ if (err < 0) {
+ goto out_err;
+ }
err = v9fs_co_remove(pdu->s, &fidp->path);
if (!err) {
err = offset;
}
-
+out_err:
/* For TREMOVE we need to clunk the fid even on failed remove */
clunk_fid(pdu->s, fidp->fid);
put_fid(pdu->s, fidp);
@@ -2823,3 +2993,14 @@ void handle_9p_output(VirtIODevice *vdev, VirtQueue *vq)
}
free_pdu(s, pdu);
}
+
+void virtio_9p_set_fd_limit(void)
+{
+ struct rlimit rlim;
+ if (getrlimit(RLIMIT_NOFILE, &rlim) < 0) {
+ fprintf(stderr, "Failed to get the resource limit\n");
+ exit(1);
+ }
+ open_fd_hw = rlim.rlim_cur - MIN(400, rlim.rlim_cur/3);
+ open_fd_rc = rlim.rlim_cur/2;
+}
diff --git a/hw/9pfs/virtio-9p.h b/hw/9pfs/virtio-9p.h
index e16e5f4..28fe1ff 100644
--- a/hw/9pfs/virtio-9p.h
+++ b/hw/9pfs/virtio-9p.h
@@ -5,6 +5,7 @@
#include <dirent.h>
#include <sys/time.h>
#include <utime.h>
+#include <sys/resource.h>
#include "hw/virtio.h"
#include "fsdev/file-op-9p.h"
@@ -101,6 +102,9 @@ enum p9_proto_version {
#define P9_NOTAG (u16)(~0)
#define P9_NOFID (u32)(~0)
#define P9_MAXWELEM 16
+
+#define FID_REFERENCED 0x1
+#define FID_NON_RECLAIMABLE 0x2
static inline const char *rpath(FsContext *ctx, const char *path, char *buffer)
{
snprintf(buffer, PATH_MAX, "%s/%s", ctx->fs_root, path);
@@ -198,14 +202,21 @@ struct V9fsFidState
int32_t fid;
V9fsString path;
union {
- int fd;
- DIR *dir;
- V9fsXattr xattr;
+ int fd;
+ DIR *dir;
+ V9fsXattr xattr;
} fs;
+ union {
+ int fd;
+ DIR *dir;
+ } fs_reclaim;
+ int flags;
+ int open_flags;
uid_t uid;
int ref;
int clunked;
V9fsFidState *next;
+ V9fsFidState *rclm_lst;
};
typedef struct V9fsState
@@ -354,6 +365,9 @@ typedef struct V9fsGetlock
V9fsString client_id;
} V9fsGetlock;
+extern int open_fd_hw;
+extern int total_open_fd;
+
size_t pdu_packunpack(void *addr, struct iovec *sg, int sg_count,
size_t offset, size_t size, int pack);
@@ -364,4 +378,6 @@ static inline size_t do_pdu_unpack(void *dst, struct iovec *sg, int sg_count,
}
extern void handle_9p_output(VirtIODevice *vdev, VirtQueue *vq);
+extern void virtio_9p_set_fd_limit(void);
+extern void v9fs_reclaim_fd(V9fsState *s);
#endif
--
1.7.4.1
^ permalink raw reply related [flat|nested] 7+ messages in thread