On Thu, Mar 29, 2018 at 6:42 PM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
> On Thu, Mar 29, 2018 at 6:58 PM, Miklos Szeredi <miklos@xxxxxxxxxx> wrote:
>> On Thu, Mar 29, 2018 at 4:18 PM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
>>> On 64bit systems, when overlay layers are not all on the same fs, but
>>> all inode numbers of underlying fs are not using the high bits, use the
>>> high bits to partition the overlay st_ino address space. The high bits
>>> hold the fsid (upper fsid is 0). This way overlay inode numbers are unique
>>> and all inodes use overlay st_dev. Inode numbers are also persistent
>>> for a given layer configuration.
>>>
>>> Currently, our only indication for available high ino bits is from a
>>> filesystem that supports file handles and uses the default encode_fh()
>>> operation, which encodes a 32bit inode number.
>>>
>>> Signed-off-by: Amir Goldstein <amir73il@xxxxxxxxx>
>>> ---
>>> fs/overlayfs/inode.c | 31 +++++++++++++++++++++++++++++--
>>> fs/overlayfs/overlayfs.h | 3 ++-
>>> fs/overlayfs/ovl_entry.h | 2 ++
>>> fs/overlayfs/super.c | 26 ++++++++++++++++++++++----
>>> fs/overlayfs/util.c | 24 +++++++++++++++++++++---
>>> 5 files changed, 76 insertions(+), 10 deletions(-)
>>>
>>> diff --git a/fs/overlayfs/inode.c b/fs/overlayfs/inode.c
>>> index 89dfab20fe0e..7fc9c83bf2ff 100644
>>> --- a/fs/overlayfs/inode.c
>>> +++ b/fs/overlayfs/inode.c
>>> @@ -63,6 +63,7 @@ static int ovl_map_dev_ino(struct dentry *dentry, struct kstat *stat)
>>> {
>>> struct ovl_layer *lower_layer = ovl_layer_lower(dentry);
>>> bool samefs = ovl_same_sb(dentry->d_sb);
>>> + int xinobits = ovl_xino_bits(dentry->d_sb);
>>>
>>> if (samefs) {
>>> /*
>>> @@ -71,7 +72,31 @@ static int ovl_map_dev_ino(struct dentry *dentry, struct kstat *stat)
>>> * which is friendly to du -x.
>>> */
>>> stat->dev = dentry->d_sb->s_dev;
>>> - } else if (S_ISDIR(dentry->d_inode->i_mode)) {
>>> + return 0;
>>> + } else if (xinobits) {
>>> + /*
>>> + * All inode numbers of underlying fs should not be using the
>>> + * high xinobits, so we use high xinobits to partition the
>>> + * overlay st_ino address space. The high bits holds the fsid
>>> + * (upper fsid is 0). This way overlay inode numbers are unique
>>> + * and all inodes use overlay st_dev. Inode numbers are also
>>> + * persistent for a given layer configuration.
>>> + */
>>> + if (stat->ino >> (64 - xinobits)) {
>>> + pr_warn_ratelimited("overlayfs: inode number too big (%pd2, ino=%llu, xinobits=%d)\n",
>>> + dentry, stat->ino, xinobits);
>>> + } else {
>>> + if (lower_layer) {
>>> + stat->ino |= ((u64)lower_layer->fsid)
>>> + << (64 - xinobits);
>>> + }
>>> + stat->dev = dentry->d_sb->s_dev;
>>> + return 0;
>>> + }
>>> + }
>>> +
>>> + /* The inode could not be mapped to a unified st_ino address space */
>>> + if (S_ISDIR(dentry->d_inode->i_mode)) {
>>> /*
>>> * Always use the overlay st_dev for directories, so 'find
>>> * -xdev' will scan the entire overlay mount and won't cross the
>>> @@ -117,11 +142,13 @@ int ovl_getattr(const struct path *path, struct kstat *stat,
>>> /*
>>> * For non-dir or same fs, we use st_ino of the copy up origin.
>>> * This guaranties constant st_dev/st_ino across copy up.
>>> + * With xino feature and non-samefs, we use st_ino of the copy up
>>> + * origin masked with high bits that represent the layer id.
>>> *
>>> * If lower filesystem supports NFS file handles, this also guaranties
>>> * persistent st_ino across mount cycle.
>>> */
>>> - if (!is_dir || samefs) {
>>> + if (!is_dir || samefs || ovl_xino_bits(dentry->d_sb)) {
>>> if (OVL_TYPE_ORIGIN(type)) {
>>> struct kstat lowerstat;
>>> u32 lowermask = STATX_INO | (!is_dir ? STATX_NLINK : 0);
>>> diff --git a/fs/overlayfs/overlayfs.h b/fs/overlayfs/overlayfs.h
>>> index 4432599ecbc6..09f5b8ce70aa 100644
>>> --- a/fs/overlayfs/overlayfs.h
>>> +++ b/fs/overlayfs/overlayfs.h
>>> @@ -202,7 +202,8 @@ void ovl_drop_write(struct dentry *dentry);
>>> struct dentry *ovl_workdir(struct dentry *dentry);
>>> const struct cred *ovl_override_creds(struct super_block *sb);
>>> struct super_block *ovl_same_sb(struct super_block *sb);
>>> -bool ovl_can_decode_fh(struct super_block *sb);
>>> +int ovl_xino_bits(struct super_block *sb);
>>> +int ovl_can_decode_fh(struct super_block *sb);
>>> struct dentry *ovl_indexdir(struct super_block *sb);
>>> bool ovl_index_all(struct super_block *sb);
>>> bool ovl_verify_lower(struct super_block *sb);
>>> diff --git a/fs/overlayfs/ovl_entry.h b/fs/overlayfs/ovl_entry.h
>>> index e1c838c27a74..6a077fb2a75f 100644
>>> --- a/fs/overlayfs/ovl_entry.h
>>> +++ b/fs/overlayfs/ovl_entry.h
>>> @@ -63,6 +63,8 @@ struct ovl_fs {
>>> /* Did we take the inuse lock? */
>>> bool upperdir_locked;
>>> bool workdir_locked;
>>> + /* Inode numbers in all layers do not use the high xino_bits */
>>> + int xino_bits;
>>> };
>>>
>>> /* private information held for every overlayfs dentry */
>>> diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
>>> index 7d97d30cad39..d7284444f404 100644
>>> --- a/fs/overlayfs/super.c
>>> +++ b/fs/overlayfs/super.c
>>> @@ -17,6 +17,7 @@
>>> #include <linux/statfs.h>
>>> #include <linux/seq_file.h>
>>> #include <linux/posix_acl_xattr.h>
>>> +#include <linux/exportfs.h>
>>> #include "overlayfs.h"
>>>
>>> MODULE_AUTHOR("Miklos Szeredi <miklos@xxxxxxxxxx>");
>>> @@ -701,6 +702,7 @@ static int ovl_check_namelen(struct path *path, struct ovl_fs *ofs,
>>> static int ovl_lower_dir(const char *name, struct path *path,
>>> struct ovl_fs *ofs, int *stack_depth, bool *remote)
>>> {
>>> + int fh_type;
>>> int err;
>>>
>>> err = ovl_mount_dir_noesc(name, path);
>>> @@ -720,15 +722,19 @@ static int ovl_lower_dir(const char *name, struct path *path,
>>> * The inodes index feature and NFS export need to encode and decode
>>> * file handles, so they require that all layers support them.
>>> */
>>> + fh_type = ovl_can_decode_fh(path->dentry->d_sb);
>>> if ((ofs->config.nfs_export ||
>>> - (ofs->config.index && ofs->config.upperdir)) &&
>>> - !ovl_can_decode_fh(path->dentry->d_sb)) {
>>> + (ofs->config.index && ofs->config.upperdir)) && !fh_type) {
>>> ofs->config.index = false;
>>> ofs->config.nfs_export = false;
>>> pr_warn("overlayfs: fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n",
>>> name);
>>> }
>>>
>>> + /* Check if lower fs has 32bit inode numbers */
>>> + if (fh_type != FILEID_INO32_GEN)
>>> + ofs->xino_bits = 0;
>>> +
>>> return 0;
>>>
>>> out_put:
>>> @@ -952,6 +958,7 @@ static int ovl_make_workdir(struct ovl_fs *ofs, struct path *workpath)
>>> {
>>> struct vfsmount *mnt = ofs->upper_mnt;
>>> struct dentry *temp;
>>> + int fh_type;
>>> int err;
>>>
>>> err = mnt_want_write(mnt);
>>> @@ -1001,12 +1008,16 @@ static int ovl_make_workdir(struct ovl_fs *ofs, struct path *workpath)
>>> }
>>>
>>> /* Check if upper/work fs supports file handles */
>>> - if (ofs->config.index &&
>>> - !ovl_can_decode_fh(ofs->workdir->d_sb)) {
>>> + fh_type = ovl_can_decode_fh(ofs->workdir->d_sb);
>>> + if (ofs->config.index && !fh_type) {
>>> ofs->config.index = false;
>>> pr_warn("overlayfs: upper fs does not support file handles, falling back to index=off.\n");
>>> }
>>>
>>> + /* Check if upper fs has 32bit inode numbers */
>>> + if (fh_type != FILEID_INO32_GEN)
>>> + ofs->xino_bits = 0;
>>> +
>>> /* NFS export of r/w mount depends on index */
>>> if (ofs->config.nfs_export && !ofs->config.index) {
>>> pr_warn("overlayfs: NFS export requires \"index=on\", falling back to nfs_export=off.\n");
>>> @@ -1185,6 +1196,11 @@ static int ovl_get_lower_layers(struct ovl_fs *ofs, struct path *stack,
>>> }
>>> ofs->numlower++;
>>> }
>>> +
>>> + /* When all layers on same fs, overlay can use real inode numbers */
>>> + if (!ofs->numlowerfs || (ofs->numlowerfs == 1 && !ofs->upper_mnt))
>>> + ofs->xino_bits = 0;
>>> +
>>> err = 0;
>>> out:
>>> return err;
>>> @@ -1308,6 +1324,8 @@ static int ovl_fill_super(struct super_block *sb, void *data, int silent)
>>>
>>> sb->s_stack_depth = 0;
>>> sb->s_maxbytes = MAX_LFS_FILESIZE;
>>> + /* Assume underlaying fs uses 32bit inodes unless proven otherwise */
>>> + ofs->xino_bits = BITS_PER_LONG - 32;
>>
>> This disables xino for 32bit archs. Which is probably the right thing
>> to do, otherwise there might be a regression in some cases since
>> kernel will return EOVERFLOW if st_ino would overflow. Well, this is
>> true for 32bit mode in 64bit kernel as well, so the above is not a
>> perfect solution.
>>
>> Not sure if we need to worry. For 32bit archs, I think disabling xino
>> is OK; it can be enabled explicitly if needed. For 64bit archs, let's
>> hope it doesn't regress for anybody and if it does, we need to take
>> steps.
>
> Perhaps, the steps would be to implement -o noxino.
And that would need to be the default if selected by a kernel config
option. Should we do that now, to prevent surprises? Maybe we
should; it does't cost anything in complexity, and I think it's better
to let the user choose between backward compatibility and new feature.
Thanks,
Miklos
--
To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
