[Devel] [PATCH RH7] ipc: add export_operations to mqueue sb
Andrey Zhadchenko
andrey.zhadchenko at virtuozzo.com
Thu Sep 24 16:18:15 MSK 2020
For some reason i have forgotten to send a copy to devel at openvz.org
during this conversation
Highlits:
>>>> +
>>>> +#define MQUEUE_INODE_HASH_BITS 7
>>>
>>> Why you've chosen smaller hash size than for cgroups?
>>> #define CGROUP_INODE_HASH_BITS 10
>>>
>>
>> I think inotify on mqueue mount is quite rare. It doesn't look like
>> something default as inotify on cgroup for kubernetes.
>
>OK.
>>>> + mqueue_hash_del(inode);
>>>> +
>>>
>>> Looks like for cgroups we should've also put cgroup_hash_del to
>>> ->evict_inode helper, to be able to use igrab in
>>> cgroup_find_inode.
>>> Shouldn't we?
>>>
>>
>> This is very good idea, looks more robust. I should make a patch.
>
>Nice!
On Thu, 24 Sep 2020 13:38:24 +0300
Konstantin Khorenko <khorenko at virtuozzo.com> wrote:
> On 09/09/2020 01:19 PM, Pavel Tikhomirov wrote:
> > Looks good.
> >
> > Reviewed-by: Pavel Tikhomirov <ptikhomirov at virtuozzo.com>
> >
> > But please also see two non-critical questions below.
>
> Andrey, any answer on that?
>
>
> > On 9/8/20 5:33 PM, Andrey Zhadchenko wrote:
> >> criu uses fhandle from fdinfo to dump inotify objects, but mqueue
> >> has no export_operations in super block. However, they are needed
> >> if we want to operate with files using fhandles.
> >> Add hash table as a storage for exported inodes.
> >>
> >> https://jira.sw.ru/browse/PSBM-103727
> >> Signed-off-by: Andrey Zhadchenko <andrey.zhadchenko at virtuozzo.com>
> >> ---
> >> ipc/mqueue.c | 150
> >> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file
> >> changed, 150 insertions(+)
> >>
> >> diff --git a/ipc/mqueue.c b/ipc/mqueue.c
> >> index 458e655..48ec6f8 100644
> >> --- a/ipc/mqueue.c
> >> +++ b/ipc/mqueue.c
> >> @@ -35,6 +35,8 @@
> >> #include <linux/ipc_namespace.h>
> >> #include <linux/user_namespace.h>
> >> #include <linux/slab.h>
> >> +#include <linux/exportfs.h>
> >> +#include <linux/random.h>
> >>
> >> #include <net/sock.h>
> >> #include "util.h"
> >> @@ -86,6 +88,7 @@ struct mqueue_inode_info {
> >> };
> >>
> >> static const struct inode_operations mqueue_dir_inode_operations;
> >> +static const struct export_operations mqueue_export_ops;
> >> static const struct file_operations mqueue_file_operations;
> >> static const struct super_operations mqueue_super_ops;
> >> static void remove_notification(struct mqueue_inode_info *info);
> >> @@ -117,6 +120,144 @@ static struct ipc_namespace
> >> *get_ns_from_inode(struct inode *inode) return ns;
> >> }
> >>
> >> +/*
> >> + * hashtable for inodes that have exported fhandles.
> >> + * When we export fhandle, we add it's inode into
> >> + * hashtable so we can find it fast
> >> + */
> >> +
> >> +#define MQUEUE_INODE_HASH_BITS 7
> >
> > Why you've chosen smaller hash size than for cgroups?
> > #define CGROUP_INODE_HASH_BITS 10
> >
> >> +static DEFINE_HASHTABLE(mqueue_inode_table,
> >> MQUEUE_INODE_HASH_BITS); +static
> >> DEFINE_SPINLOCK(mqueue_inode_table_lock); +
> >> +struct mq_inode_hitem {
> >> + struct inode *inode;
> >> + struct hlist_node hlist;
> >> +};
> >> +
> >> +static inline unsigned long mqueue_inode_get_hash(unsigned int
> >> i_ino) +{
> >> + return hash_32(i_ino, MQUEUE_INODE_HASH_BITS);
> >> +}
> >> +
> >> +static struct mq_inode_hitem *mqueue_find_item_no_lock(unsigned
> >> long fh[2]) +{
> >> + struct mq_inode_hitem *i;
> >> + struct hlist_head *head = mqueue_inode_table
> >> + + mqueue_inode_get_hash(fh[1]);
> >> + struct mq_inode_hitem *found = NULL;
> >> +
> >> + hlist_for_each_entry(i, head, hlist) {
> >> + if (i->inode->i_generation == fh[0] &&
> >> + i->inode->i_ino == fh[1]) {
> >> + found = i;
> >> + break;
> >> + }
> >> + }
> >> +
> >> + return found;
> >> +}
> >> +
> >> +static struct inode *mqueue_find_inode(unsigned long fh[2], char
> >> take_ref) +{
> >> + struct mq_inode_hitem *item;
> >> + struct inode *ret = NULL;
> >> +
> >> + spin_lock(&mqueue_inode_table_lock);
> >> + item = mqueue_find_item_no_lock(fh);
> >> + if (item && take_ref && !igrab(item->inode))
> >> + item = NULL;
> >> + spin_unlock(&mqueue_inode_table_lock);
> >> +
> >> + if (item)
> >> + ret = item->inode;
> >> +
> >> + return ret;
> >> +}
> >> +
> >> +static int mqueue_hash_add(struct inode *inode)
> >> +{
> >> + unsigned long fh[2] = {inode->i_generation, inode->i_ino};
> >> +
> >> + if (!mqueue_find_inode(fh, 0)) {
> >> + struct mq_inode_hitem *item;
> >> + struct mq_inode_hitem *existing_item = 0;
> >> + struct hlist_head *head = mqueue_inode_table
> >> + + mqueue_inode_get_hash(inode->i_ino);
> >> +
> >> + item = kmalloc(sizeof(struct mq_inode_hitem),
> >> GFP_KERNEL);
> >> + if (!item)
> >> + return -ENOMEM;
> >> + item->inode = inode;
> >> +
> >> + spin_lock(&mqueue_inode_table_lock);
> >> + existing_item = mqueue_find_item_no_lock(fh);
> >> + if (!existing_item)
> >> + hlist_add_head(&item->hlist, head);
> >> + spin_unlock(&mqueue_inode_table_lock);
> >> +
> >> + if (existing_item)
> >> + kfree(item);
> >> + }
> >> +
> >> + return 0;
> >> +}
> >> +
> >> +static void mqueue_hash_del(struct inode *inode)
> >> +{
> >> + struct mq_inode_hitem *item;
> >> + unsigned long fh[2] = {inode->i_generation, inode->i_ino};
> >> +
> >> + spin_lock(&mqueue_inode_table_lock);
> >> + item = mqueue_find_item_no_lock(fh);
> >> + if (item)
> >> + hlist_del(&item->hlist);
> >> + spin_unlock(&mqueue_inode_table_lock);
> >> +
> >> + kfree(item);
> >> + return;
> >> +}
> >> +
> >> +static struct dentry *mqueue_fh_to_dentry(struct super_block *sb,
> >> + struct fid *fid, int fh_len, int fh_type)
> >> +{
> >> + struct inode *inode;
> >> + struct dentry *dentry = ERR_PTR(-ENOENT);
> >> + unsigned long fhandle[2] = {fid->raw[0], fid->raw[1]};
> >> +
> >> + if (fh_len < 2)
> >> + return NULL;
> >> +
> >> + inode = mqueue_find_inode(fhandle, 1);
> >> + if (inode) {
> >> + dentry = d_find_alias(inode);
> >> + iput(inode);
> >> + }
> >> + return dentry;
> >> +}
> >> +
> >> +static int mqueue_encode_fh(struct inode *inode, __u32 *fh, int
> >> *len,
> >> + struct inode *parent)
> >> +{
> >> + if (*len < 2) {
> >> + *len = 2;
> >> + return FILEID_INVALID;
> >> + }
> >> +
> >> + /*
> >> + * encode_fh is expected to return 255 (FILEID_INVALID)
> >> + * in case of failure. We can't return ENOMEM, so
> >> + * return FILEID_INVALID at least.
> >> + */
> >> + if (mqueue_hash_add(inode))
> >> + return FILEID_INVALID;
> >> +
> >> + fh[0] = inode->i_generation;
> >> + fh[1] = inode->i_ino;
> >> + *len = 2;
> >> + return 1;
> >> +}
> >> +
> >> /* Auxiliary functions to manipulate messages' list */
> >> static int msg_insert(struct msg_msg *msg, struct
> >> mqueue_inode_info *info) {
> >> @@ -226,6 +367,7 @@ static struct inode *mqueue_get_inode(struct
> >> super_block *sb, goto err;
> >>
> >> inode->i_ino = get_next_ino();
> >> + inode->i_generation = prandom_u32();
> >> inode->i_mode = mode;
> >> inode->i_uid = current_fsuid();
> >> inode->i_gid = current_fsgid();
> >> @@ -316,6 +458,7 @@ static int mqueue_fill_super(struct
> >> super_block *sb, void *data, int silent) sb->s_blocksize_bits =
> >> PAGE_CACHE_SHIFT; sb->s_magic = MQUEUE_MAGIC;
> >> sb->s_op = &mqueue_super_ops;
> >> + sb->s_export_op = &mqueue_export_ops;
> >>
> >> inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX |
> >> S_IRWXUGO, NULL); if (IS_ERR(inode))
> >> @@ -382,6 +525,8 @@ static void mqueue_evict_inode(struct inode
> >> *inode) if (S_ISDIR(inode->i_mode))
> >> return;
> >>
> >> + mqueue_hash_del(inode);
> >> +
> >
> > Looks like for cgroups we should've also put cgroup_hash_del to
> > ->evict_inode helper, to be able to use igrab in
> > cgroup_find_inode. Shouldn't we?
> >
> >> ipc_ns = get_ns_from_inode(inode);
> >> info = MQUEUE_I(inode);
> >> spin_lock(&info->lock);
> >> @@ -1376,6 +1521,11 @@ static const struct inode_operations
> >> mqueue_dir_inode_operations = { .unlink = mqueue_unlink,
> >> };
> >>
> >> +static const struct export_operations mqueue_export_ops = {
> >> + .encode_fh = mqueue_encode_fh,
> >> + .fh_to_dentry = mqueue_fh_to_dentry,
> >> +};
> >> +
> >> static const struct file_operations mqueue_file_operations = {
> >> .flush = mqueue_flush_file,
> >> .poll = mqueue_poll_file,
> >>
> >
More information about the Devel
mailing list