[PATCH 4/5] Check for feature flags needed to read from the pool.

[Massimo Maggi]

Ported to GRUB2 from IllumOS changeset 13700    2889e2596bd6

Author: Christopher Siden <address@hidden>
Ported-by: Massimo Maggi <address@hidden>
---
 grub-core/fs/zfs/zfs.c | 102
++++++++++++++++++++++++++++++++++++++++++++++++-
 include/grub/zfs/dmu.h |  46 +++++++++++++++++++++-
 include/grub/zfs/zfs.h |   1 +
 3 files changed, 146 insertions(+), 3 deletions(-)

diff --git a/grub-core/fs/zfs/zfs.c b/grub-core/fs/zfs/zfs.c
index 7f76445..cf10534 100644
--- a/grub-core/fs/zfs/zfs.c
+++ b/grub-core/fs/zfs/zfs.c
@@ -272,6 +272,19 @@ grub_crypto_cipher_handle_t (*grub_zfs_load_key)
(const struct grub_zfs_key *key
                                                  grub_size_t keysize,
                                                  grub_uint64_t salt,
                                                  grub_uint64_t algo) = NULL;
+/*
+ * List of pool features that the grub implementation of ZFS supports for
+ * read. Note that features that are only required for write do not need
+ * to be listed here since grub opens pools in read-only mode.
+ */
+static const char *spa_feature_names[] = {
+       "max.test:feat1",NULL
+};
+
+static int
+check_feature(const char *name, grub_uint64_t val, struct
grub_zfs_dir_ctx *ctx);
+static int
+check_mos_features(dnode_phys_t *mosmdn_phys,grub_zfs_endian_t
endian,struct grub_zfs_data* data );
  static grub_err_t  zlib_decompress (void *s, void *d,
@@ -927,7 +940,7 @@ check_pool_label (struct grub_zfs_data *data,
                  int *inserted)
 {
   grub_uint64_t pool_state, txg = 0;
-  char *nvlist;
+  char *nvlist,*features;
 #if 0
   char *nv;
 #endif
@@ -1047,7 +1060,30 @@ check_pool_label (struct grub_zfs_data *data,
     grub_free (nv);
   }
   grub_dprintf ("zfs", "check 10 passed\n");
-
+  if ((features=grub_zfs_nvlist_lookup_nvlist(nvlist,
ZPOOL_CONFIG_FEATURES_FOR_READ)))
+  {
+    const char *nvp=NULL;
+    char *name = grub_zalloc(51);
+    char *nameptr;
+    int namelen;
+    while ((nvp = nvlist_next_nvpair(features, nvp)) != NULL)
+    {
+      nvpair_name(nvp, &nameptr,&namelen);
+      if(namelen > 50){namelen=50;}
+      grub_strncpy(name,nameptr,namelen);
+      name[namelen]=0;
+      grub_dprintf("zfs","namelen=%u str=%s\n",namelen,name);
+      if (check_feature(name,1, NULL) != 0)
+      {
+       grub_dprintf("zfs","feature missing in check_pool_label:%s\n",name);
+       err= grub_error (GRUB_ERR_NOT_IMPLEMENTED_YET," check_pool_label
missing feature '%s' for read",name);
+       grub_free(name);
+       return err;
+      }
+    }
+    grub_free(name);
+  }
+  grub_dprintf ("zfs", "check 12 passed (feature flags)\n");
   grub_free (nvlist);
    return GRUB_ERR_NONE;
@@ -3502,6 +3538,10 @@ zfs_mount (grub_device_t dev)
       return NULL;
     }
 +    if (ub->ub_version >= SPA_VERSION_FEATURES &&
+       check_mos_features(&((objset_phys_t *) osp)->os_meta_dnode,ub_endian,
data) != 0)
+         return NULL;
+       
   /* Got the MOS. Save it at the memory addr MOS. */
   grub_memmove (&(data->mos.dn), &((objset_phys_t *) osp)->os_meta_dnode,
                DNODE_SIZE);
@@ -4042,6 +4082,64 @@ grub_zfs_dir (grub_device_t device, const char *path,
   return grub_errno;
 }
 +static int
+check_feature(const char *name, grub_uint64_t
val,__attribute__((unused)) struct grub_zfs_dir_ctx *ctx)
+{
+  int i;
+  if(val ==0) return 0;
+  if(*name==0) return 0;
+  for (i = 0; spa_feature_names[i] != NULL; i++) +  {
+    if (grub_strcmp(name, spa_feature_names[i]) == 0) +        return 0;
+  }
+  grub_printf("missing feature for read '%s'\n",name);
+  return 1;
+}
+
+/*
+ * Checks whether the MOS features that are active are supported by this
+ * (GRUB's) implementation of ZFS.
+ *
+ * Return:
+ *     0: Success.
+ *     errnum: Failure.
+ */
+                  +static int
+check_mos_features(dnode_phys_t *mosmdn_phys,grub_zfs_endian_t
endian,struct grub_zfs_data* data )
+{
+  grub_uint64_t objnum;
+  grub_uint8_t errnum = 0;
+  dnode_end_t dn,mosmdn;
+  mzap_phys_t* mzp;
+  grub_zfs_endian_t endianzap;
+  int size;
+  grub_memmove(&(mosmdn.dn),mosmdn_phys,sizeof(dnode_phys_t));
+  mosmdn.endian=endian;
+  if ((errnum = dnode_get(&mosmdn, DMU_POOL_DIRECTORY_OBJECT,
+         DMU_OT_OBJECT_DIRECTORY, &dn,data)) != 0)
+      return (errnum);
+
+  /*
+   * Find the object number for 'features_for_read' and retrieve its
+   * corresponding dnode. Note that we don't check features_for_write
+   * because GRUB is not opening the pool for write.
+   */
+  if ((errnum = zap_lookup(&dn, DMU_POOL_FEATURES_FOR_READ, &objnum,
data,0)) != 0)
+      return (errnum);
+  +  if ((errnum = dnode_get(&mosmdn, objnum, DMU_OTN_ZAP_METADATA,
&dn, data)) != 0)
+      return (errnum);
+
+  if ((errnum = dmu_read(&dn, 0, (void**)&mzp, &endianzap,data)) != 0)
+      return (errnum);
+
+  size = grub_zfs_to_cpu16 (dn.dn.dn_datablkszsec, dn.endian) <<
SPA_MINBLOCKSHIFT;
+  return (mzap_iterate(mzp,endianzap, size, check_feature,NULL));
+}
+
+
 #ifdef GRUB_UTIL
 static grub_err_t
 grub_zfs_embed (grub_device_t device __attribute__ ((unused)),
diff --git a/include/grub/zfs/dmu.h b/include/grub/zfs/dmu.h
index 8fc6dc5..4ad616c 100644
--- a/include/grub/zfs/dmu.h
+++ b/include/grub/zfs/dmu.h
@@ -22,6 +22,39 @@
  #ifndef       _SYS_DMU_H
 #define        _SYS_DMU_H
+#define        B_FALSE 0
+#define        B_TRUE  1
+
+#define        DMU_OT_NEWTYPE 0x80
+#define        DMU_OT_METADATA 0x40
+#define        DMU_OT_BYTESWAP_MASK 0x3f
+
+#define        DMU_OT(byteswap, metadata) \
+       (DMU_OT_NEWTYPE | \
+       ((metadata) ? DMU_OT_METADATA : 0) | \
+       ((byteswap) & DMU_OT_BYTESWAP_MASK))
+
+#define        DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
+       ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
+       (ot) < DMU_OT_NUMTYPES)
+
+#define        DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
+       ((ot) & DMU_OT_METADATA) : \
+       dmu_ot[(ot)].ot_metadata)
+
+typedef enum dmu_object_byteswap {
+       DMU_BSWAP_UINT8,
+       DMU_BSWAP_UINT16,
+       DMU_BSWAP_UINT32,
+       DMU_BSWAP_UINT64,
+       DMU_BSWAP_ZAP,
+       DMU_BSWAP_DNODE,
+       DMU_BSWAP_OBJSET,
+       DMU_BSWAP_ZNODE,
+       DMU_BSWAP_OLDACL,
+       DMU_BSWAP_ACL,
+       DMU_BSWAP_NUMFUNCS
+} dmu_object_byteswap_t;
  /*
  * This file describes the interface that the DMU provides for its
@@ -89,7 +122,17 @@ typedef enum dmu_object_type {
        DMU_OT_SA_ATTR_REGISTRATION,    /* ZAP */
        DMU_OT_SA_ATTR_LAYOUTS,         /* ZAP */
        DMU_OT_DSL_KEYCHAIN = 54,
-       DMU_OT_NUMTYPES
+       DMU_OT_NUMTYPES,
+       DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE),
+       DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE),
+       DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE),
+       DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE),
+       DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE),
+       DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE),
+       DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE),
+       DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE),
+       DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE),
+       DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE),
 } dmu_object_type_t;
  typedef enum dmu_objset_type {
@@ -116,5 +159,6 @@ typedef enum dmu_objset_type {
 #define        DMU_POOL_HISTORY                "history"
 #define        DMU_POOL_PROPS                  "pool_props"
 #define        DMU_POOL_L2CACHE                "l2cache"
+#define        DMU_POOL_FEATURES_FOR_READ      "features_for_read"
  #endif        /* _SYS_DMU_H */
diff --git a/include/grub/zfs/zfs.h b/include/grub/zfs/zfs.h
index 20b23b7..761ade7 100644
--- a/include/grub/zfs/zfs.h
+++ b/include/grub/zfs/zfs.h
@@ -81,6 +81,7 @@ typedef enum grub_zfs_endian
 #define        ZPOOL_CONFIG_DDT_HISTOGRAM      "ddt_histogram"
 #define        ZPOOL_CONFIG_DDT_OBJ_STATS      "ddt_object_stats"
 #define        ZPOOL_CONFIG_DDT_STATS          "ddt_stats"
+#define        ZPOOL_CONFIG_FEATURES_FOR_READ  "features_for_read"
 /*
  * The persistent vdev state is stored as separate values rather than a
single
  * 'vdev_state' entry.  This is because a device can be in multiple
states, such
-- 
1.8.1.1

Attached Message Part
Description: Text document

signature.asc
Description: OpenPGP digital signature