[SRU][F][PATCH 0/1] ceph: fix inode number handling on arches with 32-bit ino_t (LP: 1899582)

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

[SRU][F][PATCH 0/1] ceph: fix inode number handling on arches with 32-bit ino_t (LP: 1899582)

frank.heimes
BugLink: https://bugs.launchpad.net/bugs/1899582

SRU Justification:

[Impact]

* A problem occurs with 32-bit ino_t values on 64-bit architectures (like s390x).

* Without this patch, cephfs kernel client does not work properly on s390x: the initial symtom is missing the first created file/directory on cephfs mount point as reported in https://tracker.ceph.com/issues/46828.

[Fix]

* cherry-pick for groovy: ebce3eb2f7ef9f6ef01a60874ebd232450107c9a ebce3eb2f7ef "ceph: fix inode number handling on arches with 32-bit ino_t"

* backport for focal: https://launchpadlibrarian.net/501822049/0001-ceph-fix-inode-number-handling-on-arches-with-32-bit.patch

[Test Case]

* IBM Development team and Test team have carried out simple testing and stress testing by automatically creating many Ceph clusters on s390x (with different configuration each time), doing the cephfs kernel mounting, checking inode numbers, running stress tests. The same procedures were done on x86_64 arch by Test team.

[Regression Potential]

* There is a certain regression risk with this patch, since:

* common ceph filesystem code in fs/ceph/ is modified

* the inode handling is touched

* ceph is used for supplemental volumes

* and as the original author of this patch (Jeff Layton) wrote in the commit message, the potential issue might happen on 32-bit arches only and he also provided workaround for those arches.

[Other]

* The above patch/commit was upstream accepted with kernel 5.9.

* Cherry picking the patch/commit to groovy works fine, hence this SRU.

* For focal a cherry pick does not apply cleanly, hence a backport got added to LP 1899582.

Tuan Hoang (1):
  ceph: fix inode number handling on arches with 32-bit ino_t

 fs/ceph/caps.c    | 12 ++++----
 fs/ceph/debugfs.c |  2 +-
 fs/ceph/dir.c     | 29 ++++++++-----------
 fs/ceph/inode.c   | 21 +++++++-------
 fs/ceph/quota.c   |  4 +--
 fs/ceph/super.h   | 73 ++++++++++++++++++++++++-----------------------
 6 files changed, 69 insertions(+), 72 deletions(-)

--
2.25.1


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

[SRU][F][PATCH 1/1] ceph: fix inode number handling on arches with 32-bit ino_t

frank.heimes
From: Tuan Hoang <[hidden email]>

BugLink: https://bugs.launchpad.net/bugs/1899582

Upstream commit : ebce3eb2f7ef9f6ef01a60874ebd232450107c9a

Tuan and Ulrich mentioned that they were hitting a problem on s390x,
which has a 32-bit ino_t value, even though it's a 64-bit arch (for
historical reasons).

I think the current handling of inode numbers in the ceph driver is
wrong. It tries to use 32-bit inode numbers on 32-bit arches, but that's
actually not a problem. 32-bit arches can deal with 64-bit inode numbers
just fine when userland code is compiled with LFS support (the common
case these days).

What we really want to do is just use 64-bit numbers everywhere, unless
someone has mounted with the ino32 mount option. In that case, we want
to ensure that we hash the inode number down to something that will fit
in 32 bits before presenting the value to userland.

Add new helper functions that do this, and only do the conversion before
presenting these values to userland in getattr and readdir.

The inode table hashvalue is changed to just cast the inode number to
unsigned long, as low-order bits are the most likely to vary anyway.

While it's not strictly required, we do want to put something in
inode->i_ino. Instead of basing it on BITS_PER_LONG, however, base it on
the size of the ino_t type.

NOTE: This is a user-visible change on 32-bit arches:

1/ inode numbers will be seen to have changed between kernel versions.
   32-bit arches will see large inode numbers now instead of the hashed
   ones they saw before.

2/ any really old software not built with LFS support may start failing
   stat() calls with -EOVERFLOW on inode numbers >2^32. Nothing much we
   can do about these, but hopefully the intersection of people running
   such code on ceph will be very small.

The workaround for both problems is to mount with "-o ino32".

[ idryomov: changelog tweak ]

URL: https://tracker.ceph.com/issues/46828
Reported-by: Ulrich Weigand <[hidden email]>
Reported-and-Tested-by: Tuan Hoang1 <[hidden email]>
Signed-off-by: Jeff Layton <[hidden email]>
Reviewed-by: "Yan, Zheng" <[hidden email]>
Signed-off-by: Ilya Dryomov <[hidden email]>
(backported from commit ebce3eb2f7ef9f6ef01a60874ebd232450107c9a)
Signed-off-by: Frank Heimes <[hidden email]>
---
 fs/ceph/caps.c    | 12 ++++----
 fs/ceph/debugfs.c |  2 +-
 fs/ceph/dir.c     | 29 ++++++++-----------
 fs/ceph/inode.c   | 21 +++++++-------
 fs/ceph/quota.c   |  4 +--
 fs/ceph/super.h   | 73 ++++++++++++++++++++++++-----------------------
 6 files changed, 69 insertions(+), 72 deletions(-)

diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index b2695919435e..785e0ebe9890 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -869,8 +869,8 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch)
  int have = ci->i_snap_caps;
 
  if ((have & mask) == mask) {
- dout("__ceph_caps_issued_mask ino 0x%lx snap issued %s"
-     " (mask %s)\n", ci->vfs_inode.i_ino,
+ dout("__ceph_caps_issued_mask ino 0x%llx snap issued %s"
+     " (mask %s)\n", ceph_ino(&ci->vfs_inode),
      ceph_cap_string(have),
      ceph_cap_string(mask));
  return 1;
@@ -881,8 +881,8 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch)
  if (!__cap_is_valid(cap))
  continue;
  if ((cap->issued & mask) == mask) {
- dout("__ceph_caps_issued_mask ino 0x%lx cap %p issued %s"
-     " (mask %s)\n", ci->vfs_inode.i_ino, cap,
+ dout("__ceph_caps_issued_mask ino 0x%llx cap %p issued %s"
+     " (mask %s)\n", ceph_ino(&ci->vfs_inode), cap,
      ceph_cap_string(cap->issued),
      ceph_cap_string(mask));
  if (touch)
@@ -893,8 +893,8 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch)
  /* does a combination of caps satisfy mask? */
  have |= cap->issued;
  if ((have & mask) == mask) {
- dout("__ceph_caps_issued_mask ino 0x%lx combo issued %s"
-     " (mask %s)\n", ci->vfs_inode.i_ino,
+ dout("__ceph_caps_issued_mask ino 0x%llx combo issued %s"
+     " (mask %s)\n", ceph_ino(&ci->vfs_inode),
      ceph_cap_string(cap->issued),
      ceph_cap_string(mask));
  if (touch) {
diff --git a/fs/ceph/debugfs.c b/fs/ceph/debugfs.c
index facb387c2735..712b43322547 100644
--- a/fs/ceph/debugfs.c
+++ b/fs/ceph/debugfs.c
@@ -128,7 +128,7 @@ static int caps_show_cb(struct inode *inode, struct ceph_cap *cap, void *p)
 {
  struct seq_file *s = p;
 
- seq_printf(s, "0x%-17lx%-17s%-17s\n", inode->i_ino,
+ seq_printf(s, "0x%-17llx%-17s%-17s\n", ceph_ino(inode),
    ceph_cap_string(cap->issued),
    ceph_cap_string(cap->implemented));
  return 0;
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 3367a8194f24..d8c6068429fb 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -254,9 +254,7 @@ static int __dcache_readdir(struct file *file,  struct dir_context *ctx,
      dentry, dentry, d_inode(dentry));
  ctx->pos = di->offset;
  if (!dir_emit(ctx, dentry->d_name.name,
-      dentry->d_name.len,
-      ceph_translate_ino(dentry->d_sb,
- d_inode(dentry)->i_ino),
+      dentry->d_name.len, ceph_present_inode(d_inode(dentry)),
       d_inode(dentry)->i_mode >> 12)) {
  dput(dentry);
  err = 0;
@@ -319,18 +317,21 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
  /* always start with . and .. */
  if (ctx->pos == 0) {
  dout("readdir off 0 -> '.'\n");
- if (!dir_emit(ctx, ".", 1,
-    ceph_translate_ino(inode->i_sb, inode->i_ino),
+ if (!dir_emit(ctx, ".", 1, ceph_present_inode(inode),
     inode->i_mode >> 12))
  return 0;
  ctx->pos = 1;
  }
  if (ctx->pos == 1) {
- ino_t ino = parent_ino(file->f_path.dentry);
+ u64 ino;
+ struct dentry *dentry = file->f_path.dentry;
+
+ spin_lock(&dentry->d_lock);
+ ino = ceph_present_inode(dentry->d_parent->d_inode);
+ spin_unlock(&dentry->d_lock);
+
  dout("readdir off 1 -> '..'\n");
- if (!dir_emit(ctx, "..", 2,
-    ceph_translate_ino(inode->i_sb, ino),
-    inode->i_mode >> 12))
+ if (!dir_emit(ctx, "..", 2, ino, inode->i_mode >> 12))
  return 0;
  ctx->pos = 2;
  }
@@ -498,9 +499,6 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
  }
  for (; i < rinfo->dir_nr; i++) {
  struct ceph_mds_reply_dir_entry *rde = rinfo->dir_entries + i;
- struct ceph_vino vino;
- ino_t ino;
- u32 ftype;
 
  BUG_ON(rde->offset < ctx->pos);
 
@@ -510,13 +508,10 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
      rde->name_len, rde->name, &rde->inode.in);
 
  BUG_ON(!rde->inode.in);
- ftype = le32_to_cpu(rde->inode.in->mode) >> 12;
- vino.ino = le64_to_cpu(rde->inode.in->ino);
- vino.snap = le64_to_cpu(rde->inode.in->snapid);
- ino = ceph_vino_to_ino(vino);
 
  if (!dir_emit(ctx, rde->name, rde->name_len,
-      ceph_translate_ino(inode->i_sb, ino), ftype)) {
+      ceph_present_ino(inode->i_sb, le64_to_cpu(rde->inode.in->ino)),
+      le32_to_cpu(rde->inode.in->mode) >> 12)) {
  dout("filldir stopping us...\n");
  return 0;
  }
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index c07407586ce8..7164bc8f9115 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -41,8 +41,10 @@ static void ceph_inode_work(struct work_struct *work);
  */
 static int ceph_set_ino_cb(struct inode *inode, void *data)
 {
- ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
- inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
+ struct ceph_inode_info *ci = ceph_inode(inode);
+
+ ci->i_vino = *(struct ceph_vino *)data;
+ inode->i_ino = ceph_vino_to_ino_t(ci->i_vino);
  inode_set_iversion_raw(inode, 0);
  return 0;
 }
@@ -50,19 +52,16 @@ static int ceph_set_ino_cb(struct inode *inode, void *data)
 struct inode *ceph_get_inode(struct super_block *sb, struct ceph_vino vino)
 {
  struct inode *inode;
- ino_t t = ceph_vino_to_ino(vino);
 
- inode = iget5_locked(sb, t, ceph_ino_compare, ceph_set_ino_cb, &vino);
+ inode = iget5_locked(sb, (unsigned long)vino.ino, ceph_ino_compare,
+     ceph_set_ino_cb, &vino);
  if (!inode)
  return ERR_PTR(-ENOMEM);
- if (inode->i_state & I_NEW) {
- dout("get_inode created new inode %p %llx.%llx ino %llx\n",
-     inode, ceph_vinop(inode), (u64)inode->i_ino);
+ if (inode->i_state & I_NEW)
  unlock_new_inode(inode);
- }
 
- dout("get_inode on %lu=%llx.%llx got %p\n", inode->i_ino, vino.ino,
-     vino.snap, inode);
+ dout("get_inode on %llu=%llx.%llx got %p new %d\n", ceph_present_inode(inode),
+     ceph_vinop(inode), inode, !!(inode->i_state & I_NEW));
  return inode;
 }
 
@@ -2343,7 +2342,7 @@ int ceph_getattr(const struct path *path, struct kstat *stat,
  }
 
  generic_fillattr(inode, stat);
- stat->ino = ceph_translate_ino(inode->i_sb, inode->i_ino);
+ stat->ino = ceph_present_inode(inode);
 
  /*
  * btime on newly-allocated inodes is 0, so if this is still set to
diff --git a/fs/ceph/quota.c b/fs/ceph/quota.c
index 19507e2fdb57..bda26b4f9b97 100644
--- a/fs/ceph/quota.c
+++ b/fs/ceph/quota.c
@@ -23,12 +23,12 @@ static inline bool ceph_has_realms_with_quotas(struct inode *inode)
 {
  struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
  struct super_block *sb = mdsc->fsc->sb;
+ struct inode *root = d_inode(sb->s_root);
 
  if (atomic64_read(&mdsc->quotarealms_count) > 0)
  return true;
  /* if root is the real CephFS root, we don't have quota realms */
- if (sb->s_root->d_inode &&
-    (sb->s_root->d_inode->i_ino == CEPH_INO_ROOT))
+ if (root && ceph_ino(root) == CEPH_INO_ROOT)
  return false;
  /* otherwise, we can't know for sure */
  return true;
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index bb12c9f3a218..fa7cd81cf232 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -427,15 +427,7 @@ static inline struct ceph_vino ceph_vino(struct inode *inode)
  return ceph_inode(inode)->i_vino;
 }
 
-/*
- * ino_t is <64 bits on many architectures, blech.
- *
- *               i_ino (kernel inode)   st_ino (userspace)
- * i386          32                     32
- * x86_64+ino32  64                     32
- * x86_64        64                     64
- */
-static inline u32 ceph_ino_to_ino32(__u64 vino)
+static inline u32 ceph_ino_to_ino32(u64 vino)
 {
  u32 ino = vino & 0xffffffff;
  ino ^= vino >> 32;
@@ -445,34 +437,17 @@ static inline u32 ceph_ino_to_ino32(__u64 vino)
 }
 
 /*
- * kernel i_ino value
+ * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
+ * some arches. We generally do not use this value inside the ceph driver, but
+ * we do want to set it to something, so that generic vfs code has an
+ * appropriate value for tracepoints and the like.
  */
-static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
+static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
 {
-#if BITS_PER_LONG == 32
- return ceph_ino_to_ino32(vino.ino);
-#else
+ if (sizeof(ino_t) == sizeof(u32))
+ return ceph_ino_to_ino32(vino.ino);
  return (ino_t)vino.ino;
-#endif
-}
-
-/*
- * user-visible ino (stat, filldir)
- */
-#if BITS_PER_LONG == 32
-static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
-{
- return ino;
-}
-#else
-static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
-{
- if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
- ino = ceph_ino_to_ino32(ino);
- return ino;
 }
-#endif
-
 
 /* for printf-style formatting */
 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
@@ -481,11 +456,34 @@ static inline u64 ceph_ino(struct inode *inode)
 {
  return ceph_inode(inode)->i_vino.ino;
 }
+
 static inline u64 ceph_snap(struct inode *inode)
 {
  return ceph_inode(inode)->i_vino.snap;
 }
 
+/**
+ * ceph_present_ino - format an inode number for presentation to userland
+ * @sb: superblock where the inode lives
+ * @ino: inode number to (possibly) convert
+ *
+ * If the user mounted with the ino32 option, then the 64-bit value needs
+ * to be converted to something that can fit inside 32 bits. Note that
+ * internal kernel code never uses this value, so this is entirely for
+ * userland consumption.
+ */
+static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
+{
+ if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
+ return ceph_ino_to_ino32(ino);
+ return ino;
+}
+
+static inline u64 ceph_present_inode(struct inode *inode)
+{
+ return ceph_present_ino(inode->i_sb, ceph_ino(inode));
+}
+
 static inline int ceph_ino_compare(struct inode *inode, void *data)
 {
  struct ceph_vino *pvino = (struct ceph_vino *)data;
@@ -494,11 +492,16 @@ static inline int ceph_ino_compare(struct inode *inode, void *data)
  ci->i_vino.snap == pvino->snap;
 }
 
+
 static inline struct inode *ceph_find_inode(struct super_block *sb,
     struct ceph_vino vino)
 {
- ino_t t = ceph_vino_to_ino(vino);
- return ilookup5(sb, t, ceph_ino_compare, &vino);
+ /*
+ * NB: The hashval will be run through the fs/inode.c hash function
+ * anyway, so there is no need to squash the inode number down to
+ * 32-bits first. Just use low-order bits on arches with 32-bit long.
+ */
+ return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
 }
 
 
--
2.25.1


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

ACK/Cmnt: [SRU][F][PATCH 1/1] ceph: fix inode number handling on arches with 32-bit ino_t

Stefan Bader-2
On 14.10.20 21:12, [hidden email] wrote:

> From: Tuan Hoang <[hidden email]>
>
> BugLink: https://bugs.launchpad.net/bugs/1899582
>
> Upstream commit : ebce3eb2f7ef9f6ef01a60874ebd232450107c9a
>
> Tuan and Ulrich mentioned that they were hitting a problem on s390x,
> which has a 32-bit ino_t value, even though it's a 64-bit arch (for
> historical reasons).
>
> I think the current handling of inode numbers in the ceph driver is
> wrong. It tries to use 32-bit inode numbers on 32-bit arches, but that's
> actually not a problem. 32-bit arches can deal with 64-bit inode numbers
> just fine when userland code is compiled with LFS support (the common
> case these days).
>
> What we really want to do is just use 64-bit numbers everywhere, unless
> someone has mounted with the ino32 mount option. In that case, we want
> to ensure that we hash the inode number down to something that will fit
> in 32 bits before presenting the value to userland.
>
> Add new helper functions that do this, and only do the conversion before
> presenting these values to userland in getattr and readdir.
>
> The inode table hashvalue is changed to just cast the inode number to
> unsigned long, as low-order bits are the most likely to vary anyway.
>
> While it's not strictly required, we do want to put something in
> inode->i_ino. Instead of basing it on BITS_PER_LONG, however, base it on
> the size of the ino_t type.
>
> NOTE: This is a user-visible change on 32-bit arches:
>
> 1/ inode numbers will be seen to have changed between kernel versions.
>    32-bit arches will see large inode numbers now instead of the hashed
>    ones they saw before.
>
> 2/ any really old software not built with LFS support may start failing
>    stat() calls with -EOVERFLOW on inode numbers >2^32. Nothing much we
>    can do about these, but hopefully the intersection of people running
>    such code on ceph will be very small.
>
> The workaround for both problems is to mount with "-o ino32".
>
> [ idryomov: changelog tweak ]
>
> URL: https://tracker.ceph.com/issues/46828
> Reported-by: Ulrich Weigand <[hidden email]>
> Reported-and-Tested-by: Tuan Hoang1 <[hidden email]>
> Signed-off-by: Jeff Layton <[hidden email]>
> Reviewed-by: "Yan, Zheng" <[hidden email]>
> Signed-off-by: Ilya Dryomov <[hidden email]>
> (backported from commit ebce3eb2f7ef9f6ef01a60874ebd232450107c9a)
> Signed-off-by: Frank Heimes <[hidden email]>
Acked-by: Stefan Bader <[hidden email]>
> ---

Upstream backport which appears to be tested well with good results. In Focal
the only 32bit arch left is AFAIK armhf. Work-around for the user-space visible
change is possible.

What could have been added is a note about how much had to be done for the
backport (like context adjustments, function renames, ...).

-Stefan

>  fs/ceph/caps.c    | 12 ++++----
>  fs/ceph/debugfs.c |  2 +-
>  fs/ceph/dir.c     | 29 ++++++++-----------
>  fs/ceph/inode.c   | 21 +++++++-------
>  fs/ceph/quota.c   |  4 +--
>  fs/ceph/super.h   | 73 ++++++++++++++++++++++++-----------------------
>  6 files changed, 69 insertions(+), 72 deletions(-)
>
> diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
> index b2695919435e..785e0ebe9890 100644
> --- a/fs/ceph/caps.c
> +++ b/fs/ceph/caps.c
> @@ -869,8 +869,8 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch)
>   int have = ci->i_snap_caps;
>  
>   if ((have & mask) == mask) {
> - dout("__ceph_caps_issued_mask ino 0x%lx snap issued %s"
> -     " (mask %s)\n", ci->vfs_inode.i_ino,
> + dout("__ceph_caps_issued_mask ino 0x%llx snap issued %s"
> +     " (mask %s)\n", ceph_ino(&ci->vfs_inode),
>       ceph_cap_string(have),
>       ceph_cap_string(mask));
>   return 1;
> @@ -881,8 +881,8 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch)
>   if (!__cap_is_valid(cap))
>   continue;
>   if ((cap->issued & mask) == mask) {
> - dout("__ceph_caps_issued_mask ino 0x%lx cap %p issued %s"
> -     " (mask %s)\n", ci->vfs_inode.i_ino, cap,
> + dout("__ceph_caps_issued_mask ino 0x%llx cap %p issued %s"
> +     " (mask %s)\n", ceph_ino(&ci->vfs_inode), cap,
>       ceph_cap_string(cap->issued),
>       ceph_cap_string(mask));
>   if (touch)
> @@ -893,8 +893,8 @@ int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch)
>   /* does a combination of caps satisfy mask? */
>   have |= cap->issued;
>   if ((have & mask) == mask) {
> - dout("__ceph_caps_issued_mask ino 0x%lx combo issued %s"
> -     " (mask %s)\n", ci->vfs_inode.i_ino,
> + dout("__ceph_caps_issued_mask ino 0x%llx combo issued %s"
> +     " (mask %s)\n", ceph_ino(&ci->vfs_inode),
>       ceph_cap_string(cap->issued),
>       ceph_cap_string(mask));
>   if (touch) {
> diff --git a/fs/ceph/debugfs.c b/fs/ceph/debugfs.c
> index facb387c2735..712b43322547 100644
> --- a/fs/ceph/debugfs.c
> +++ b/fs/ceph/debugfs.c
> @@ -128,7 +128,7 @@ static int caps_show_cb(struct inode *inode, struct ceph_cap *cap, void *p)
>  {
>   struct seq_file *s = p;
>  
> - seq_printf(s, "0x%-17lx%-17s%-17s\n", inode->i_ino,
> + seq_printf(s, "0x%-17llx%-17s%-17s\n", ceph_ino(inode),
>     ceph_cap_string(cap->issued),
>     ceph_cap_string(cap->implemented));
>   return 0;
> diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
> index 3367a8194f24..d8c6068429fb 100644
> --- a/fs/ceph/dir.c
> +++ b/fs/ceph/dir.c
> @@ -254,9 +254,7 @@ static int __dcache_readdir(struct file *file,  struct dir_context *ctx,
>       dentry, dentry, d_inode(dentry));
>   ctx->pos = di->offset;
>   if (!dir_emit(ctx, dentry->d_name.name,
> -      dentry->d_name.len,
> -      ceph_translate_ino(dentry->d_sb,
> - d_inode(dentry)->i_ino),
> +      dentry->d_name.len, ceph_present_inode(d_inode(dentry)),
>        d_inode(dentry)->i_mode >> 12)) {
>   dput(dentry);
>   err = 0;
> @@ -319,18 +317,21 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
>   /* always start with . and .. */
>   if (ctx->pos == 0) {
>   dout("readdir off 0 -> '.'\n");
> - if (!dir_emit(ctx, ".", 1,
> -    ceph_translate_ino(inode->i_sb, inode->i_ino),
> + if (!dir_emit(ctx, ".", 1, ceph_present_inode(inode),
>      inode->i_mode >> 12))
>   return 0;
>   ctx->pos = 1;
>   }
>   if (ctx->pos == 1) {
> - ino_t ino = parent_ino(file->f_path.dentry);
> + u64 ino;
> + struct dentry *dentry = file->f_path.dentry;
> +
> + spin_lock(&dentry->d_lock);
> + ino = ceph_present_inode(dentry->d_parent->d_inode);
> + spin_unlock(&dentry->d_lock);
> +
>   dout("readdir off 1 -> '..'\n");
> - if (!dir_emit(ctx, "..", 2,
> -    ceph_translate_ino(inode->i_sb, ino),
> -    inode->i_mode >> 12))
> + if (!dir_emit(ctx, "..", 2, ino, inode->i_mode >> 12))
>   return 0;
>   ctx->pos = 2;
>   }
> @@ -498,9 +499,6 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
>   }
>   for (; i < rinfo->dir_nr; i++) {
>   struct ceph_mds_reply_dir_entry *rde = rinfo->dir_entries + i;
> - struct ceph_vino vino;
> - ino_t ino;
> - u32 ftype;
>  
>   BUG_ON(rde->offset < ctx->pos);
>  
> @@ -510,13 +508,10 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
>       rde->name_len, rde->name, &rde->inode.in);
>  
>   BUG_ON(!rde->inode.in);
> - ftype = le32_to_cpu(rde->inode.in->mode) >> 12;
> - vino.ino = le64_to_cpu(rde->inode.in->ino);
> - vino.snap = le64_to_cpu(rde->inode.in->snapid);
> - ino = ceph_vino_to_ino(vino);
>  
>   if (!dir_emit(ctx, rde->name, rde->name_len,
> -      ceph_translate_ino(inode->i_sb, ino), ftype)) {
> +      ceph_present_ino(inode->i_sb, le64_to_cpu(rde->inode.in->ino)),
> +      le32_to_cpu(rde->inode.in->mode) >> 12)) {
>   dout("filldir stopping us...\n");
>   return 0;
>   }
> diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
> index c07407586ce8..7164bc8f9115 100644
> --- a/fs/ceph/inode.c
> +++ b/fs/ceph/inode.c
> @@ -41,8 +41,10 @@ static void ceph_inode_work(struct work_struct *work);
>   */
>  static int ceph_set_ino_cb(struct inode *inode, void *data)
>  {
> - ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
> - inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
> + struct ceph_inode_info *ci = ceph_inode(inode);
> +
> + ci->i_vino = *(struct ceph_vino *)data;
> + inode->i_ino = ceph_vino_to_ino_t(ci->i_vino);
>   inode_set_iversion_raw(inode, 0);
>   return 0;
>  }
> @@ -50,19 +52,16 @@ static int ceph_set_ino_cb(struct inode *inode, void *data)
>  struct inode *ceph_get_inode(struct super_block *sb, struct ceph_vino vino)
>  {
>   struct inode *inode;
> - ino_t t = ceph_vino_to_ino(vino);
>  
> - inode = iget5_locked(sb, t, ceph_ino_compare, ceph_set_ino_cb, &vino);
> + inode = iget5_locked(sb, (unsigned long)vino.ino, ceph_ino_compare,
> +     ceph_set_ino_cb, &vino);
>   if (!inode)
>   return ERR_PTR(-ENOMEM);
> - if (inode->i_state & I_NEW) {
> - dout("get_inode created new inode %p %llx.%llx ino %llx\n",
> -     inode, ceph_vinop(inode), (u64)inode->i_ino);
> + if (inode->i_state & I_NEW)
>   unlock_new_inode(inode);
> - }
>  
> - dout("get_inode on %lu=%llx.%llx got %p\n", inode->i_ino, vino.ino,
> -     vino.snap, inode);
> + dout("get_inode on %llu=%llx.%llx got %p new %d\n", ceph_present_inode(inode),
> +     ceph_vinop(inode), inode, !!(inode->i_state & I_NEW));
>   return inode;
>  }
>  
> @@ -2343,7 +2342,7 @@ int ceph_getattr(const struct path *path, struct kstat *stat,
>   }
>  
>   generic_fillattr(inode, stat);
> - stat->ino = ceph_translate_ino(inode->i_sb, inode->i_ino);
> + stat->ino = ceph_present_inode(inode);
>  
>   /*
>   * btime on newly-allocated inodes is 0, so if this is still set to
> diff --git a/fs/ceph/quota.c b/fs/ceph/quota.c
> index 19507e2fdb57..bda26b4f9b97 100644
> --- a/fs/ceph/quota.c
> +++ b/fs/ceph/quota.c
> @@ -23,12 +23,12 @@ static inline bool ceph_has_realms_with_quotas(struct inode *inode)
>  {
>   struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
>   struct super_block *sb = mdsc->fsc->sb;
> + struct inode *root = d_inode(sb->s_root);
>  
>   if (atomic64_read(&mdsc->quotarealms_count) > 0)
>   return true;
>   /* if root is the real CephFS root, we don't have quota realms */
> - if (sb->s_root->d_inode &&
> -    (sb->s_root->d_inode->i_ino == CEPH_INO_ROOT))
> + if (root && ceph_ino(root) == CEPH_INO_ROOT)
>   return false;
>   /* otherwise, we can't know for sure */
>   return true;
> diff --git a/fs/ceph/super.h b/fs/ceph/super.h
> index bb12c9f3a218..fa7cd81cf232 100644
> --- a/fs/ceph/super.h
> +++ b/fs/ceph/super.h
> @@ -427,15 +427,7 @@ static inline struct ceph_vino ceph_vino(struct inode *inode)
>   return ceph_inode(inode)->i_vino;
>  }
>  
> -/*
> - * ino_t is <64 bits on many architectures, blech.
> - *
> - *               i_ino (kernel inode)   st_ino (userspace)
> - * i386          32                     32
> - * x86_64+ino32  64                     32
> - * x86_64        64                     64
> - */
> -static inline u32 ceph_ino_to_ino32(__u64 vino)
> +static inline u32 ceph_ino_to_ino32(u64 vino)
>  {
>   u32 ino = vino & 0xffffffff;
>   ino ^= vino >> 32;
> @@ -445,34 +437,17 @@ static inline u32 ceph_ino_to_ino32(__u64 vino)
>  }
>  
>  /*
> - * kernel i_ino value
> + * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
> + * some arches. We generally do not use this value inside the ceph driver, but
> + * we do want to set it to something, so that generic vfs code has an
> + * appropriate value for tracepoints and the like.
>   */
> -static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
> +static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
>  {
> -#if BITS_PER_LONG == 32
> - return ceph_ino_to_ino32(vino.ino);
> -#else
> + if (sizeof(ino_t) == sizeof(u32))
> + return ceph_ino_to_ino32(vino.ino);
>   return (ino_t)vino.ino;
> -#endif
> -}
> -
> -/*
> - * user-visible ino (stat, filldir)
> - */
> -#if BITS_PER_LONG == 32
> -static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
> -{
> - return ino;
> -}
> -#else
> -static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
> -{
> - if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
> - ino = ceph_ino_to_ino32(ino);
> - return ino;
>  }
> -#endif
> -
>  
>  /* for printf-style formatting */
>  #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
> @@ -481,11 +456,34 @@ static inline u64 ceph_ino(struct inode *inode)
>  {
>   return ceph_inode(inode)->i_vino.ino;
>  }
> +
>  static inline u64 ceph_snap(struct inode *inode)
>  {
>   return ceph_inode(inode)->i_vino.snap;
>  }
>  
> +/**
> + * ceph_present_ino - format an inode number for presentation to userland
> + * @sb: superblock where the inode lives
> + * @ino: inode number to (possibly) convert
> + *
> + * If the user mounted with the ino32 option, then the 64-bit value needs
> + * to be converted to something that can fit inside 32 bits. Note that
> + * internal kernel code never uses this value, so this is entirely for
> + * userland consumption.
> + */
> +static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
> +{
> + if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
> + return ceph_ino_to_ino32(ino);
> + return ino;
> +}
> +
> +static inline u64 ceph_present_inode(struct inode *inode)
> +{
> + return ceph_present_ino(inode->i_sb, ceph_ino(inode));
> +}
> +
>  static inline int ceph_ino_compare(struct inode *inode, void *data)
>  {
>   struct ceph_vino *pvino = (struct ceph_vino *)data;
> @@ -494,11 +492,16 @@ static inline int ceph_ino_compare(struct inode *inode, void *data)
>   ci->i_vino.snap == pvino->snap;
>  }
>  
> +
>  static inline struct inode *ceph_find_inode(struct super_block *sb,
>      struct ceph_vino vino)
>  {
> - ino_t t = ceph_vino_to_ino(vino);
> - return ilookup5(sb, t, ceph_ino_compare, &vino);
> + /*
> + * NB: The hashval will be run through the fs/inode.c hash function
> + * anyway, so there is no need to squash the inode number down to
> + * 32-bits first. Just use low-order bits on arches with 32-bit long.
> + */
> + return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
>  }
>  
>  
>


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team

signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

ACK: [SRU][F][PATCH 1/1] ceph: fix inode number handling on arches with 32-bit ino_t

Andrea Righi
In reply to this post by frank.heimes
On Wed, Oct 14, 2020 at 09:12:01PM +0200, [hidden email] wrote:

> From: Tuan Hoang <[hidden email]>
>
> BugLink: https://bugs.launchpad.net/bugs/1899582
>
> Upstream commit : ebce3eb2f7ef9f6ef01a60874ebd232450107c9a
>
> Tuan and Ulrich mentioned that they were hitting a problem on s390x,
> which has a 32-bit ino_t value, even though it's a 64-bit arch (for
> historical reasons).
>
> I think the current handling of inode numbers in the ceph driver is
> wrong. It tries to use 32-bit inode numbers on 32-bit arches, but that's
> actually not a problem. 32-bit arches can deal with 64-bit inode numbers
> just fine when userland code is compiled with LFS support (the common
> case these days).
>
> What we really want to do is just use 64-bit numbers everywhere, unless
> someone has mounted with the ino32 mount option. In that case, we want
> to ensure that we hash the inode number down to something that will fit
> in 32 bits before presenting the value to userland.
>
> Add new helper functions that do this, and only do the conversion before
> presenting these values to userland in getattr and readdir.
>
> The inode table hashvalue is changed to just cast the inode number to
> unsigned long, as low-order bits are the most likely to vary anyway.
>
> While it's not strictly required, we do want to put something in
> inode->i_ino. Instead of basing it on BITS_PER_LONG, however, base it on
> the size of the ino_t type.
>
> NOTE: This is a user-visible change on 32-bit arches:
>
> 1/ inode numbers will be seen to have changed between kernel versions.
>    32-bit arches will see large inode numbers now instead of the hashed
>    ones they saw before.
>
> 2/ any really old software not built with LFS support may start failing
>    stat() calls with -EOVERFLOW on inode numbers >2^32. Nothing much we
>    can do about these, but hopefully the intersection of people running
>    such code on ceph will be very small.
>
> The workaround for both problems is to mount with "-o ino32".
>
> [ idryomov: changelog tweak ]
>
> URL: https://tracker.ceph.com/issues/46828
> Reported-by: Ulrich Weigand <[hidden email]>
> Reported-and-Tested-by: Tuan Hoang1 <[hidden email]>
> Signed-off-by: Jeff Layton <[hidden email]>
> Reviewed-by: "Yan, Zheng" <[hidden email]>
> Signed-off-by: Ilya Dryomov <[hidden email]>
> (backported from commit ebce3eb2f7ef9f6ef01a60874ebd232450107c9a)
> Signed-off-by: Frank Heimes <[hidden email]>

As already mentioned by Stefan, you could have mentioned some details
about the backport activity (in particular about the struct cap_wait
part that is not needed in focal). But the fix makes sense and the
backport looks good to me, thanks!

Acked-by: Andrea Righi <[hidden email]>

--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team