Commit a33f6432 authored by Yan, Zheng's avatar Yan, Zheng Committed by Ilya Dryomov

ceph: encode inodes' parent/d_name in cap reconnect message

Since nautilus, MDS tracks dirfrags whose child inodes have caps in open
file table. When MDS recovers, it prefetches all of these dirfrags. This
avoids using backtrace to load inodes. But dirfrags prefetch may load
lots of useless inodes into cache, and make MDS run out of memory.

Recent MDS adds an option that disables dirfrags prefetch. When dirfrags
prefetch is disabled. Recovering MDS only prefetches corresponding dir
inodes. Including inodes' parent/d_name in cap reconnect message can
help MDS to load inodes into its cache.
Signed-off-by: default avatar"Yan, Zheng" <zyan@redhat.com>
Reviewed-by: default avatarJeff Layton <jlayton@kernel.org>
Signed-off-by: default avatarIlya Dryomov <idryomov@gmail.com>
parent bbf5c979
...@@ -3612,6 +3612,39 @@ static int send_reconnect_partial(struct ceph_reconnect_state *recon_state) ...@@ -3612,6 +3612,39 @@ static int send_reconnect_partial(struct ceph_reconnect_state *recon_state)
return err; return err;
} }
static struct dentry* d_find_primary(struct inode *inode)
{
struct dentry *alias, *dn = NULL;
if (hlist_empty(&inode->i_dentry))
return NULL;
spin_lock(&inode->i_lock);
if (hlist_empty(&inode->i_dentry))
goto out_unlock;
if (S_ISDIR(inode->i_mode)) {
alias = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
if (!IS_ROOT(alias))
dn = dget(alias);
goto out_unlock;
}
hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
spin_lock(&alias->d_lock);
if (!d_unhashed(alias) &&
(ceph_dentry(alias)->flags & CEPH_DENTRY_PRIMARY_LINK)) {
dn = dget_dlock(alias);
}
spin_unlock(&alias->d_lock);
if (dn)
break;
}
out_unlock:
spin_unlock(&inode->i_lock);
return dn;
}
/* /*
* Encode information about a cap for a reconnect with the MDS. * Encode information about a cap for a reconnect with the MDS.
*/ */
...@@ -3625,13 +3658,32 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap, ...@@ -3625,13 +3658,32 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
struct ceph_inode_info *ci = cap->ci; struct ceph_inode_info *ci = cap->ci;
struct ceph_reconnect_state *recon_state = arg; struct ceph_reconnect_state *recon_state = arg;
struct ceph_pagelist *pagelist = recon_state->pagelist; struct ceph_pagelist *pagelist = recon_state->pagelist;
int err; struct dentry *dentry;
char *path;
int pathlen, err;
u64 pathbase;
u64 snap_follows; u64 snap_follows;
dout(" adding %p ino %llx.%llx cap %p %lld %s\n", dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
inode, ceph_vinop(inode), cap, cap->cap_id, inode, ceph_vinop(inode), cap, cap->cap_id,
ceph_cap_string(cap->issued)); ceph_cap_string(cap->issued));
dentry = d_find_primary(inode);
if (dentry) {
/* set pathbase to parent dir when msg_version >= 2 */
path = ceph_mdsc_build_path(dentry, &pathlen, &pathbase,
recon_state->msg_version >= 2);
dput(dentry);
if (IS_ERR(path)) {
err = PTR_ERR(path);
goto out_err;
}
} else {
path = NULL;
pathlen = 0;
pathbase = 0;
}
spin_lock(&ci->i_ceph_lock); spin_lock(&ci->i_ceph_lock);
cap->seq = 0; /* reset cap seq */ cap->seq = 0; /* reset cap seq */
cap->issue_seq = 0; /* and issue_seq */ cap->issue_seq = 0; /* and issue_seq */
...@@ -3652,7 +3704,7 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap, ...@@ -3652,7 +3704,7 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci)); rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
rec.v2.issued = cpu_to_le32(cap->issued); rec.v2.issued = cpu_to_le32(cap->issued);
rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino); rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
rec.v2.pathbase = 0; rec.v2.pathbase = cpu_to_le64(pathbase);
rec.v2.flock_len = (__force __le32) rec.v2.flock_len = (__force __le32)
((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1); ((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);
} else { } else {
...@@ -3663,7 +3715,7 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap, ...@@ -3663,7 +3715,7 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime); ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime);
ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime); ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime);
rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino); rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
rec.v1.pathbase = 0; rec.v1.pathbase = cpu_to_le64(pathbase);
} }
if (list_empty(&ci->i_cap_snaps)) { if (list_empty(&ci->i_cap_snaps)) {
...@@ -3725,7 +3777,7 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap, ...@@ -3725,7 +3777,7 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
sizeof(struct ceph_filelock); sizeof(struct ceph_filelock);
rec.v2.flock_len = cpu_to_le32(struct_len); rec.v2.flock_len = cpu_to_le32(struct_len);
struct_len += sizeof(u32) + sizeof(rec.v2); struct_len += sizeof(u32) + pathlen + sizeof(rec.v2);
if (struct_v >= 2) if (struct_v >= 2)
struct_len += sizeof(u64); /* snap_follows */ struct_len += sizeof(u64); /* snap_follows */
...@@ -3749,7 +3801,7 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap, ...@@ -3749,7 +3801,7 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
ceph_pagelist_encode_8(pagelist, 1); ceph_pagelist_encode_8(pagelist, 1);
ceph_pagelist_encode_32(pagelist, struct_len); ceph_pagelist_encode_32(pagelist, struct_len);
} }
ceph_pagelist_encode_string(pagelist, NULL, 0); ceph_pagelist_encode_string(pagelist, path, pathlen);
ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2)); ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2));
ceph_locks_to_pagelist(flocks, pagelist, ceph_locks_to_pagelist(flocks, pagelist,
num_fcntl_locks, num_flock_locks); num_fcntl_locks, num_flock_locks);
...@@ -3758,39 +3810,20 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap, ...@@ -3758,39 +3810,20 @@ static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
out_freeflocks: out_freeflocks:
kfree(flocks); kfree(flocks);
} else { } else {
u64 pathbase = 0;
int pathlen = 0;
char *path = NULL;
struct dentry *dentry;
dentry = d_find_alias(inode);
if (dentry) {
path = ceph_mdsc_build_path(dentry,
&pathlen, &pathbase, 0);
dput(dentry);
if (IS_ERR(path)) {
err = PTR_ERR(path);
goto out_err;
}
rec.v1.pathbase = cpu_to_le64(pathbase);
}
err = ceph_pagelist_reserve(pagelist, err = ceph_pagelist_reserve(pagelist,
sizeof(u64) + sizeof(u32) + sizeof(u64) + sizeof(u32) +
pathlen + sizeof(rec.v1)); pathlen + sizeof(rec.v1));
if (err) { if (err)
goto out_freepath; goto out_err;
}
ceph_pagelist_encode_64(pagelist, ceph_ino(inode)); ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
ceph_pagelist_encode_string(pagelist, path, pathlen); ceph_pagelist_encode_string(pagelist, path, pathlen);
ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1)); ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1));
out_freepath:
ceph_mdsc_free_path(path, pathlen);
} }
out_err: out_err:
if (err >= 0) ceph_mdsc_free_path(path, pathlen);
if (!err)
recon_state->nr_caps++; recon_state->nr_caps++;
return err; return err;
} }
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment