[PATCH 1/5] Fix reporting of RAIDZ vdevs in zfsinfo

[PATCH 1/5] Fix reporting of RAIDZ vdevs in zfsinfo

[Massimo Maggi]

[-- Attachment #1.1: Type: text/plain, Size: 2148 bytes --]

Author: Massimo Maggi <me@massimo-maggi.eu>
---
 grub-core/fs/zfs/zfsinfo.c | 20 +++++++++++++++-----
 1 file changed, 15 insertions(+), 5 deletions(-)

diff --git a/grub-core/fs/zfs/zfsinfo.c b/grub-core/fs/zfs/zfsinfo.c
index fdb587a..c96bf21 100644
--- a/grub-core/fs/zfs/zfsinfo.c
+++ b/grub-core/fs/zfs/zfsinfo.c
@@ -132,21 +132,31 @@ print_vdev_info (char *nvlist, int tab)
       grub_free (path);
       return GRUB_ERR_NONE;
     }
+  char is_mirror=(grub_strcmp(type,VDEV_TYPE_MIRROR) == 0);
+  char is_raidz=(grub_strcmp(type,VDEV_TYPE_RAIDZ) == 0);
 -  if (grub_strcmp (type, VDEV_TYPE_MIRROR) == 0)
+  if (is_mirror || is_raidz)
     {
       int nelm, i;
        nelm = grub_zfs_nvlist_lookup_nvlist_array_get_nelm
 	(nvlist, ZPOOL_CONFIG_CHILDREN);
 +      if(is_mirror){
+	 grub_puts_ (N_("This VDEV is a mirror"));
+      }
+      else if(is_raidz){
+	 grub_uint64_t parity;
+	 grub_zfs_nvlist_lookup_uint64(nvlist,"nparity",&parity);
+	 grub_printf_ (N_("This VDEV is a RAIDZ%llu\n"),(unsigned long
long)parity);
+      }
       print_tabs (tab);
       if (nelm <= 0)
 	{
-	  grub_puts_ (N_("Incorrect mirror"));
+	  grub_puts_ (N_("Incorrect VDEV"));
 	  return GRUB_ERR_NONE;
 	}
-      grub_printf_ (N_("Mirror with %d children\n"), nelm);
+      grub_printf_ (N_("VDEV with %d children\n"), nelm);
       print_state (nvlist, tab);
       for (i = 0; i < nelm; i++)
 	{
@@ -162,14 +172,14 @@ print_vdev_info (char *nvlist, int tab)
 		 total element number. And the number itself is fine,
 		 only the element isn't.
 	      */
-	      grub_printf_ (N_("Mirror element number %d isn't correct\n"), i);
+	      grub_printf_ (N_("VDEV element number %d isn't correct\n"), i);
 	      continue;
 	    }
  	  /* TRANSLATORS: it's the element carying the number %d, not
 	     total element number. This is used in enumeration
 	     "Element number 1", "Element number 2", ... */
-	  grub_printf_ (N_("Mirror element number %d:\n"), i);
+	  grub_printf_ (N_("VDEV element number %d:\n"), i);
 	  print_vdev_info (child, tab + 1);
  	  grub_free (child);
-- 
1.8.1.1



[-- Attachment #1.2: Attached Message Part --]
[-- Type: text/plain, Size: 0 bytes --]



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

Re: [PATCH 1/5] Fix reporting of RAIDZ vdevs in zfsinfo

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 2679 bytes --]

The patches look ok. The patches it's based on look like original work.
However your patches are corrupted by your mailer beyond recovery. Can
you resend them as attachements with changelog mentionning the origin?

On 04.02.2013 12:08, Massimo Maggi wrote:

> Author: Massimo Maggi <me@massimo-maggi.eu>
> ---
>  grub-core/fs/zfs/zfsinfo.c | 20 +++++++++++++++-----
>  1 file changed, 15 insertions(+), 5 deletions(-)
> 
> diff --git a/grub-core/fs/zfs/zfsinfo.c b/grub-core/fs/zfs/zfsinfo.c
> index fdb587a..c96bf21 100644
> --- a/grub-core/fs/zfs/zfsinfo.c
> +++ b/grub-core/fs/zfs/zfsinfo.c
> @@ -132,21 +132,31 @@ print_vdev_info (char *nvlist, int tab)
>        grub_free (path);
>        return GRUB_ERR_NONE;
>      }
> +  char is_mirror=(grub_strcmp(type,VDEV_TYPE_MIRROR) == 0);
> +  char is_raidz=(grub_strcmp(type,VDEV_TYPE_RAIDZ) == 0);
>  -  if (grub_strcmp (type, VDEV_TYPE_MIRROR) == 0)
> +  if (is_mirror || is_raidz)
>      {
>        int nelm, i;
>         nelm = grub_zfs_nvlist_lookup_nvlist_array_get_nelm
>  	(nvlist, ZPOOL_CONFIG_CHILDREN);
>  +      if(is_mirror){
> +	 grub_puts_ (N_("This VDEV is a mirror"));
> +      }
> +      else if(is_raidz){
> +	 grub_uint64_t parity;
> +	 grub_zfs_nvlist_lookup_uint64(nvlist,"nparity",&parity);
> +	 grub_printf_ (N_("This VDEV is a RAIDZ%llu\n"),(unsigned long
> long)parity);
> +      }
>        print_tabs (tab);
>        if (nelm <= 0)
>  	{
> -	  grub_puts_ (N_("Incorrect mirror"));
> +	  grub_puts_ (N_("Incorrect VDEV"));
>  	  return GRUB_ERR_NONE;
>  	}
> -      grub_printf_ (N_("Mirror with %d children\n"), nelm);
> +      grub_printf_ (N_("VDEV with %d children\n"), nelm);
>        print_state (nvlist, tab);
>        for (i = 0; i < nelm; i++)
>  	{
> @@ -162,14 +172,14 @@ print_vdev_info (char *nvlist, int tab)
>  		 total element number. And the number itself is fine,
>  		 only the element isn't.
>  	      */
> -	      grub_printf_ (N_("Mirror element number %d isn't correct\n"), i);
> +	      grub_printf_ (N_("VDEV element number %d isn't correct\n"), i);
>  	      continue;
>  	    }
>   	  /* TRANSLATORS: it's the element carying the number %d, not
>  	     total element number. This is used in enumeration
>  	     "Element number 1", "Element number 2", ... */
> -	  grub_printf_ (N_("Mirror element number %d:\n"), i);
> +	  grub_printf_ (N_("VDEV element number %d:\n"), i);
>  	  print_vdev_info (child, tab + 1);
>   	  grub_free (child);
> 
> 
> 
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: [PATCH 1/5] Fix reporting of RAIDZ vdevs in zfsinfo

[Massimo Maggi]

[-- Attachment #1.1: Type: text/plain, Size: 3230 bytes --]

I confirm that the patch attached to this message was written by me
without copying/porting/ other code.

A commit message for this single patch could be:

The ZFS code actually present in GRUB2 is able to correctly read from
RAID-Z vdev's, but zfsinfo reports them as unknown vdev's.
This commit fixes zfsinfo to correctly report informations about RAID-Z
vdev's.

Regards,
Massimo Maggi

On 05/05/2013 00:58, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> The patches look ok. The patches it's based on look like original work.
> However your patches are corrupted by your mailer beyond recovery. Can
> you resend them as attachements with changelog mentionning the origin?
> 
> On 04.02.2013 12:08, Massimo Maggi wrote:
> 
>> Author: Massimo Maggi <me@massimo-maggi.eu>
>> ---
>>  grub-core/fs/zfs/zfsinfo.c | 20 +++++++++++++++-----
>>  1 file changed, 15 insertions(+), 5 deletions(-)
>>
>> diff --git a/grub-core/fs/zfs/zfsinfo.c b/grub-core/fs/zfs/zfsinfo.c
>> index fdb587a..c96bf21 100644
>> --- a/grub-core/fs/zfs/zfsinfo.c
>> +++ b/grub-core/fs/zfs/zfsinfo.c
>> @@ -132,21 +132,31 @@ print_vdev_info (char *nvlist, int tab)
>>        grub_free (path);
>>        return GRUB_ERR_NONE;
>>      }
>> +  char is_mirror=(grub_strcmp(type,VDEV_TYPE_MIRROR) == 0);
>> +  char is_raidz=(grub_strcmp(type,VDEV_TYPE_RAIDZ) == 0);
>>  -  if (grub_strcmp (type, VDEV_TYPE_MIRROR) == 0)
>> +  if (is_mirror || is_raidz)
>>      {
>>        int nelm, i;
>>         nelm = grub_zfs_nvlist_lookup_nvlist_array_get_nelm
>>  	(nvlist, ZPOOL_CONFIG_CHILDREN);
>>  +      if(is_mirror){
>> +	 grub_puts_ (N_("This VDEV is a mirror"));
>> +      }
>> +      else if(is_raidz){
>> +	 grub_uint64_t parity;
>> +	 grub_zfs_nvlist_lookup_uint64(nvlist,"nparity",&parity);
>> +	 grub_printf_ (N_("This VDEV is a RAIDZ%llu\n"),(unsigned long
>> long)parity);
>> +      }
>>        print_tabs (tab);
>>        if (nelm <= 0)
>>  	{
>> -	  grub_puts_ (N_("Incorrect mirror"));
>> +	  grub_puts_ (N_("Incorrect VDEV"));
>>  	  return GRUB_ERR_NONE;
>>  	}
>> -      grub_printf_ (N_("Mirror with %d children\n"), nelm);
>> +      grub_printf_ (N_("VDEV with %d children\n"), nelm);
>>        print_state (nvlist, tab);
>>        for (i = 0; i < nelm; i++)
>>  	{
>> @@ -162,14 +172,14 @@ print_vdev_info (char *nvlist, int tab)
>>  		 total element number. And the number itself is fine,
>>  		 only the element isn't.
>>  	      */
>> -	      grub_printf_ (N_("Mirror element number %d isn't correct\n"), i);
>> +	      grub_printf_ (N_("VDEV element number %d isn't correct\n"), i);
>>  	      continue;
>>  	    }
>>   	  /* TRANSLATORS: it's the element carying the number %d, not
>>  	     total element number. This is used in enumeration
>>  	     "Element number 1", "Element number 2", ... */
>> -	  grub_printf_ (N_("Mirror element number %d:\n"), i);
>> +	  grub_printf_ (N_("VDEV element number %d:\n"), i);
>>  	  print_vdev_info (child, tab + 1);
>>   	  grub_free (child);
>>
>>
>>
>> _______________________________________________
>> Grub-devel mailing list
>> Grub-devel@gnu.org
>> https://lists.gnu.org/mailman/listinfo/grub-devel
> 
> 
> 


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.2: 0001-Fix-reporting-of-RAIDZ-vdevs-in-zfsinfo.patch --]
[-- Type: text/x-patch; name="0001-Fix-reporting-of-RAIDZ-vdevs-in-zfsinfo.patch", Size: 1981 bytes --]

diff --git a/grub-core/fs/zfs/zfsinfo.c b/grub-core/fs/zfs/zfsinfo.c
index fdb587a..c96bf21 100644
--- a/grub-core/fs/zfs/zfsinfo.c
+++ b/grub-core/fs/zfs/zfsinfo.c
@@ -132,21 +132,31 @@ print_vdev_info (char *nvlist, int tab)
       grub_free (path);
       return GRUB_ERR_NONE;
     }
+  char is_mirror=(grub_strcmp(type,VDEV_TYPE_MIRROR) == 0);
+  char is_raidz=(grub_strcmp(type,VDEV_TYPE_RAIDZ) == 0);
 
-  if (grub_strcmp (type, VDEV_TYPE_MIRROR) == 0)
+  if (is_mirror || is_raidz)
     {
       int nelm, i;
 
       nelm = grub_zfs_nvlist_lookup_nvlist_array_get_nelm
 	(nvlist, ZPOOL_CONFIG_CHILDREN);
 
+      if(is_mirror){
+	 grub_puts_ (N_("This VDEV is a mirror"));
+      }
+      else if(is_raidz){
+	 grub_uint64_t parity;
+	 grub_zfs_nvlist_lookup_uint64(nvlist,"nparity",&parity);
+	 grub_printf_ (N_("This VDEV is a RAIDZ%llu\n"),(unsigned long long)parity);
+      }
       print_tabs (tab);
       if (nelm <= 0)
 	{
-	  grub_puts_ (N_("Incorrect mirror"));
+	  grub_puts_ (N_("Incorrect VDEV"));
 	  return GRUB_ERR_NONE;
 	}
-      grub_printf_ (N_("Mirror with %d children\n"), nelm);
+      grub_printf_ (N_("VDEV with %d children\n"), nelm);
       print_state (nvlist, tab);
       for (i = 0; i < nelm; i++)
 	{
@@ -162,14 +172,14 @@ print_vdev_info (char *nvlist, int tab)
 		 total element number. And the number itself is fine,
 		 only the element isn't.
 	      */
-	      grub_printf_ (N_("Mirror element number %d isn't correct\n"), i);
+	      grub_printf_ (N_("VDEV element number %d isn't correct\n"), i);
 	      continue;
 	    }
 
 	  /* TRANSLATORS: it's the element carying the number %d, not
 	     total element number. This is used in enumeration
 	     "Element number 1", "Element number 2", ... */
-	  grub_printf_ (N_("Mirror element number %d:\n"), i);
+	  grub_printf_ (N_("VDEV element number %d:\n"), i);
 	  print_vdev_info (child, tab + 1);
 
 	  grub_free (child);

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

Re: [PATCH 1/5] Fix reporting of RAIDZ vdevs in zfsinfo

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 3452 bytes --]

On 06.05.2013 17:05, Massimo Maggi wrote:

> I confirm that the patch attached to this message was written by me
> without copying/porting/ other code.
> 
> A commit message for this single patch could be:
> 
> The ZFS code actually present in GRUB2 is able to correctly read from
> RAID-Z vdev's, but zfsinfo reports them as unknown vdev's.
> This commit fixes zfsinfo to correctly report informations about RAID-Z
> vdev's.
> 

I meant other patches as well. Can you resend them as attachements?

> Regards,
> Massimo Maggi
> 
> On 05/05/2013 00:58, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>> The patches look ok. The patches it's based on look like original work.
>> However your patches are corrupted by your mailer beyond recovery. Can
>> you resend them as attachements with changelog mentionning the origin?
>>
>> On 04.02.2013 12:08, Massimo Maggi wrote:
>>
>>> Author: Massimo Maggi <me@massimo-maggi.eu>
>>> ---
>>>  grub-core/fs/zfs/zfsinfo.c | 20 +++++++++++++++-----
>>>  1 file changed, 15 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/grub-core/fs/zfs/zfsinfo.c b/grub-core/fs/zfs/zfsinfo.c
>>> index fdb587a..c96bf21 100644
>>> --- a/grub-core/fs/zfs/zfsinfo.c
>>> +++ b/grub-core/fs/zfs/zfsinfo.c
>>> @@ -132,21 +132,31 @@ print_vdev_info (char *nvlist, int tab)
>>>        grub_free (path);
>>>        return GRUB_ERR_NONE;
>>>      }
>>> +  char is_mirror=(grub_strcmp(type,VDEV_TYPE_MIRROR) == 0);
>>> +  char is_raidz=(grub_strcmp(type,VDEV_TYPE_RAIDZ) == 0);
>>>  -  if (grub_strcmp (type, VDEV_TYPE_MIRROR) == 0)
>>> +  if (is_mirror || is_raidz)
>>>      {
>>>        int nelm, i;
>>>         nelm = grub_zfs_nvlist_lookup_nvlist_array_get_nelm
>>>  	(nvlist, ZPOOL_CONFIG_CHILDREN);
>>>  +      if(is_mirror){
>>> +	 grub_puts_ (N_("This VDEV is a mirror"));
>>> +      }
>>> +      else if(is_raidz){
>>> +	 grub_uint64_t parity;
>>> +	 grub_zfs_nvlist_lookup_uint64(nvlist,"nparity",&parity);
>>> +	 grub_printf_ (N_("This VDEV is a RAIDZ%llu\n"),(unsigned long
>>> long)parity);
>>> +      }
>>>        print_tabs (tab);
>>>        if (nelm <= 0)
>>>  	{
>>> -	  grub_puts_ (N_("Incorrect mirror"));
>>> +	  grub_puts_ (N_("Incorrect VDEV"));
>>>  	  return GRUB_ERR_NONE;
>>>  	}
>>> -      grub_printf_ (N_("Mirror with %d children\n"), nelm);
>>> +      grub_printf_ (N_("VDEV with %d children\n"), nelm);
>>>        print_state (nvlist, tab);
>>>        for (i = 0; i < nelm; i++)
>>>  	{
>>> @@ -162,14 +172,14 @@ print_vdev_info (char *nvlist, int tab)
>>>  		 total element number. And the number itself is fine,
>>>  		 only the element isn't.
>>>  	      */
>>> -	      grub_printf_ (N_("Mirror element number %d isn't correct\n"), i);
>>> +	      grub_printf_ (N_("VDEV element number %d isn't correct\n"), i);
>>>  	      continue;
>>>  	    }
>>>   	  /* TRANSLATORS: it's the element carying the number %d, not
>>>  	     total element number. This is used in enumeration
>>>  	     "Element number 1", "Element number 2", ... */
>>> -	  grub_printf_ (N_("Mirror element number %d:\n"), i);
>>> +	  grub_printf_ (N_("VDEV element number %d:\n"), i);
>>>  	  print_vdev_info (child, tab + 1);
>>>   	  grub_free (child);
>>>
>>>
>>>
>>> _______________________________________________
>>> Grub-devel mailing list
>>> Grub-devel@gnu.org
>>> https://lists.gnu.org/mailman/listinfo/grub-devel
>>
>>
>>
> 




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: [PATCH 1/5] Fix reporting of RAIDZ vdevs in zfsinfo

[Massimo Maggi]

[-- Attachment #1.1: Type: text/plain, Size: 1420 bytes --]

Nella citazione in data lun 06 mag 2013 20:10:13 CEST, Vladimir
'φ-coder/phcoder' Serbinenko ha scritto:
>
> I meant other patches as well. Can you resend them as attachements?
>
Yes, you can find them attached to this message.

However I should warn you that few days ago some users have found that
the patch number 3 causes an infinite loop in grub2-probe when probing
non-ZFS partitions.
The problem is not easily reproducible and is being tracked on:
https://github.com/maxximino/grub2/issues/1
https://bugs.gentoo.org/show_bug.cgi?id=462740

I have an hypotesis about the problem (and corresponding fix), but I
need feedback from users in order to be sure that I've catched it.
Richard Yao found a workaround for this problem by reverting part of the
patch 3 "Refactoring of function nvlist_find_value", that I've ported
from IllumOS in order to keep the code as similar as possible (to ease
ports of future ZFS features).

There is also a minor cosmetic issue, as in patch 4 I've mistakenly left
a bogus feature, that gets removed by patch 5.

When the issue about the infinite loop gets fully understood and fixed,
I'll send an updated set of patches.
What do you think about excluding the refactoring of nvlist_find_value
and keeping the codebase of ZFS in Grub2 less consistent with ZFS in
Grub1/IllumOS where AFAIK new features are primarily developed?

Regards,
Massimo Maggi

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.2: 0001-Fix-reporting-of-RAIDZ-vdevs-in-zfsinfo.patch --]
[-- Type: text/x-patch; name="0001-Fix-reporting-of-RAIDZ-vdevs-in-zfsinfo.patch", Size: 1983 bytes --]

diff --git a/grub-core/fs/zfs/zfsinfo.c b/grub-core/fs/zfs/zfsinfo.c
index fdb587a..c96bf21 100644
--- a/grub-core/fs/zfs/zfsinfo.c
+++ b/grub-core/fs/zfs/zfsinfo.c
@@ -132,21 +132,31 @@ print_vdev_info (char *nvlist, int tab)
       grub_free (path);
       return GRUB_ERR_NONE;
     }
+  char is_mirror=(grub_strcmp(type,VDEV_TYPE_MIRROR) == 0);
+  char is_raidz=(grub_strcmp(type,VDEV_TYPE_RAIDZ) == 0);
 
-  if (grub_strcmp (type, VDEV_TYPE_MIRROR) == 0)
+  if (is_mirror || is_raidz)
     {
       int nelm, i;
 
       nelm = grub_zfs_nvlist_lookup_nvlist_array_get_nelm
 	(nvlist, ZPOOL_CONFIG_CHILDREN);
 
+      if(is_mirror){
+	 grub_puts_ (N_("This VDEV is a mirror"));
+      }
+      else if(is_raidz){
+	 grub_uint64_t parity;
+	 grub_zfs_nvlist_lookup_uint64(nvlist,"nparity",&parity);
+	 grub_printf_ (N_("This VDEV is a RAIDZ%llu\n"),(unsigned long long)parity);
+      }
       print_tabs (tab);
       if (nelm <= 0)
 	{
-	  grub_puts_ (N_("Incorrect mirror"));
+	  grub_puts_ (N_("Incorrect VDEV"));
 	  return GRUB_ERR_NONE;
 	}
-      grub_printf_ (N_("Mirror with %d children\n"), nelm);
+      grub_printf_ (N_("VDEV with %d children\n"), nelm);
       print_state (nvlist, tab);
       for (i = 0; i < nelm; i++)
 	{
@@ -162,14 +172,14 @@ print_vdev_info (char *nvlist, int tab)
 		 total element number. And the number itself is fine,
 		 only the element isn't.
 	      */
-	      grub_printf_ (N_("Mirror element number %d isn't correct\n"), i);
+	      grub_printf_ (N_("VDEV element number %d isn't correct\n"), i);
 	      continue;
 	    }
 
 	  /* TRANSLATORS: it's the element carying the number %d, not
 	     total element number. This is used in enumeration
 	     "Element number 1", "Element number 2", ... */
-	  grub_printf_ (N_("Mirror element number %d:\n"), i);
+	  grub_printf_ (N_("VDEV element number %d:\n"), i);
 	  print_vdev_info (child, tab + 1);
 
 	  grub_free (child);


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.3: 0002-Accept-pool-version-5000-for-feature-flags.patch --]
[-- Type: text/x-patch; name="0002-Accept-pool-version-5000-for-feature-flags.patch", Size: 2348 bytes --]

diff --git a/grub-core/fs/zfs/zfs.c b/grub-core/fs/zfs/zfs.c
index 822d65b..3ea9c95 100644
--- a/grub-core/fs/zfs/zfs.c
+++ b/grub-core/fs/zfs/zfs.c
@@ -490,15 +490,11 @@ uberblock_verify (uberblock_phys_t * ub, grub_uint64_t offset,
 
   if (grub_zfs_to_cpu64 (uber->ub_magic, GRUB_ZFS_LITTLE_ENDIAN)
       == UBERBLOCK_MAGIC
-      && grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_LITTLE_ENDIAN) > 0 
-      && grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_LITTLE_ENDIAN)
-      <= SPA_VERSION)
-    endian = GRUB_ZFS_LITTLE_ENDIAN;
+      && SPA_VERSION_IS_SUPPORTED(grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_LITTLE_ENDIAN)))
+     endian = GRUB_ZFS_LITTLE_ENDIAN;
 
   if (grub_zfs_to_cpu64 (uber->ub_magic, GRUB_ZFS_BIG_ENDIAN) == UBERBLOCK_MAGIC
-      && grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_BIG_ENDIAN) > 0 
-      && grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_BIG_ENDIAN)
-      <= SPA_VERSION)
+      && SPA_VERSION_IS_SUPPORTED(grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_BIG_ENDIAN)))
     endian = GRUB_ZFS_BIG_ENDIAN;
 
   if (endian == GRUB_ZFS_UNKNOWN_ENDIAN)
@@ -845,13 +841,13 @@ check_pool_label (struct grub_zfs_data *data,
     }
   grub_dprintf ("zfs", "check 8 passed\n");
 
-  if (version > SPA_VERSION)
+  if (!SPA_VERSION_IS_SUPPORTED(version))
     {
       grub_free (nvlist);
       return grub_error (GRUB_ERR_NOT_IMPLEMENTED_YET,
 			 "too new version %llu > %llu",
 			 (unsigned long long) version,
-			 (unsigned long long) SPA_VERSION);
+			 (unsigned long long) SPA_VERSION_BEFORE_FEATURES);
     }
   grub_dprintf ("zfs", "check 9 passed\n");
 
diff --git a/include/grub/zfs/zfs.h b/include/grub/zfs/zfs.h
index e326c8b..20b23b7 100644
--- a/include/grub/zfs/zfs.h
+++ b/include/grub/zfs/zfs.h
@@ -36,8 +36,13 @@ typedef enum grub_zfs_endian
 /*
  * On-disk version number.
  */
-#define	SPA_VERSION			33ULL
-
+#define	SPA_VERSION_INITIAL		1ULL
+#define	SPA_VERSION_BEFORE_FEATURES	33ULL
+#define	SPA_VERSION			5000ULL
+#define	SPA_VERSION_FEATURES		5000ULL
+#define	SPA_VERSION_IS_SUPPORTED(v) \
+	(((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \
+	((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))
 /*
  * The following are configuration names used in the nvlist describing a pool's
  * configuration.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.4: 0003-Refactoring-of-function-nvlist_find_value.patch --]
[-- Type: text/x-patch; name="0003-Refactoring-of-function-nvlist_find_value.patch", Size: 8074 bytes --]

diff --git a/grub-core/fs/zfs/zfs.c b/grub-core/fs/zfs/zfs.c
index 3ea9c95..7f76445 100644
--- a/grub-core/fs/zfs/zfs.c
+++ b/grub-core/fs/zfs/zfs.c
@@ -2,6 +2,7 @@
  *  GRUB  --  GRand Unified Bootloader
  *  Copyright (C) 1999,2000,2001,2002,2003,2004,2009,2010,2011  Free Software Foundation, Inc.
  *  Copyright 2010  Sun Microsystems, Inc.
+ *  Copyright (c) 2012 by Delphix. All rights reserved.
  *
  *  GRUB is free software; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
@@ -768,6 +769,155 @@ fill_vdev_info (struct grub_zfs_data *data,
 }
 
 /*
+ * For a given XDR packed nvlist, verify the first 4 bytes and move on.
+ *
+ * An XDR packed nvlist is encoded as (comments from nvs_xdr_create) :
+ *
+ *      encoding method/host endian     (4 bytes)
+ *      nvl_version                     (4 bytes)
+ *      nvl_nvflag                      (4 bytes)
+ *	encoded nvpairs:
+ *		encoded size of the nvpair      (4 bytes)
+ *		decoded size of the nvpair      (4 bytes)
+ *		name string size                (4 bytes)
+ *		name string data                (sizeof(NV_ALIGN4(string))
+ *		data type                       (4 bytes)
+ *		# of elements in the nvpair     (4 bytes)
+ *		data
+ *      2 zero's for the last nvpair
+ *		(end of the entire list)	(8 bytes)
+ *
+ */
+
+/*
+ * The nvlist_next_nvpair() function returns a handle to the next nvpair in the
+ * list following nvpair. If nvpair is NULL, the first pair is returned. If
+ * nvpair is the last pair in the nvlist, NULL is returned.
+ */
+static const char *
+nvlist_next_nvpair(const char *nvl, const char *nvpair)
+{
+	const char *nvp;
+	int encode_size;
+	int name_len;
+	if (nvl == NULL)
+		return (NULL);
+
+	if (nvpair == NULL) {
+		/* skip over header, nvl_version and nvl_nvflag */
+		nvpair = nvl + 4 * 3;
+	} else {
+		/* skip to the next nvpair */
+		encode_size = grub_be_to_cpu32 (grub_get_unaligned32(nvpair));
+		nvpair += encode_size;
+	}
+	/* 8 bytes of 0 marks the end of the list */
+	if (*(grub_uint64_t*)nvpair == 0)
+		return (NULL);
+	/*consistency checks*/
+	if (nvpair + 4 * 3 >= nvl + VDEV_PHYS_SIZE)
+	{
+	  grub_dprintf ("zfs", "nvlist overflow\n");
+	  grub_error (GRUB_ERR_BAD_FS, "incorrect nvlist");
+	  return (NULL);
+	}
+	encode_size = grub_be_to_cpu32 (grub_get_unaligned32(nvpair));
+
+	nvp = nvpair + 4*2;
+	name_len = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp += 4;
+
+	nvp = nvp + ((name_len + 3) & ~3); // align 
+	if (nvp + 4 >= nvl + VDEV_PHYS_SIZE                        
+	    || encode_size < 0
+	    || nvp + 4 + encode_size > nvl + VDEV_PHYS_SIZE)       
+	{
+	  grub_dprintf ("zfs", "nvlist overflow\n");
+	  grub_error (GRUB_ERR_BAD_FS, "incorrect nvlist");
+	  return (NULL);
+	}
+	 /* end consistency checks */
+
+	return (nvpair);
+}
+/*
+ * This function returns 0 on success and 1 on failure. On success, a string
+ * containing the name of nvpair is saved in buf.
+ */
+static int
+nvpair_name(const char *nvp, char **buf, int* buflen)
+{
+	int len;
+
+	/* skip over encode/decode size */
+	nvp += 4 * 2;
+
+	len = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp=nvp+4;
+	
+	*buf=(char*)nvp;
+	*buflen=len;
+
+	return (0);
+}
+/*
+ * This function retrieves the value of the nvpair in the form of enumerated
+ * type data_type_t.
+ */
+static int
+nvpair_type(const char *nvp)
+{
+	int name_len, type;
+
+	/* skip over encode/decode size */
+	nvp += 4 * 2;
+
+	/* skip over name_len */
+	name_len = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp += 4;
+
+	/* skip over name */
+	nvp = nvp + ((name_len + 3) & ~3); /* align */
+
+	type = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+
+	return (type);
+}
+static int
+nvpair_value(const char *nvp,char **val,
+		   grub_size_t *size_out, grub_size_t *nelm_out)
+{
+	int name_len,nelm,encode_size;
+
+	/* skip over encode/decode size */
+	encode_size = grub_be_to_cpu32 (grub_get_unaligned32(nvp));
+	nvp += 8;
+
+	/* skip over name_len */
+	name_len = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp += 4;
+
+	/* skip over name */
+	nvp = nvp + ((name_len + 3) & ~3); /* align */
+	
+	/* skip over type */
+	nvp += 4;
+	nelm = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp +=4;
+	if (nelm < 1)
+	{
+	  grub_error (GRUB_ERR_BAD_FS, "empty nvpair");
+	  return 0;
+	}
+	  *val = (char *) nvp;
+	  *size_out = encode_size;
+	  if (nelm_out)
+	    *nelm_out = nelm;
+	    
+	return 1;
+}
+
+/*
  * Check the disk label information and retrieve needed vdev name-value pairs.
  *
  */
@@ -3064,34 +3214,14 @@ dnode_get_fullpath (const char *fullpath, struct subvolume *subvol,
   return err;
 }
 
-/*
- * For a given XDR packed nvlist, verify the first 4 bytes and move on.
- *
- * An XDR packed nvlist is encoded as (comments from nvs_xdr_create) :
- *
- *      encoding method/host endian     (4 bytes)
- *      nvl_version                     (4 bytes)
- *      nvl_nvflag                      (4 bytes)
- *	encoded nvpairs:
- *		encoded size of the nvpair      (4 bytes)
- *		decoded size of the nvpair      (4 bytes)
- *		name string size                (4 bytes)
- *		name string data                (sizeof(NV_ALIGN4(string))
- *		data type                       (4 bytes)
- *		# of elements in the nvpair     (4 bytes)
- *		data
- *      2 zero's for the last nvpair
- *		(end of the entire list)	(8 bytes)
- *
- */
-
 static int
 nvlist_find_value (const char *nvlist_in, const char *name,
 		   int valtype, char **val,
 		   grub_size_t *size_out, grub_size_t *nelm_out)
 {
-  int name_len, type, encode_size;
-  const char *nvpair, *nvp_name, *nvlist = nvlist_in;
+  int name_len, type ;
+  const char *nvpair=NULL,*nvlist=nvlist_in;
+  char *nvp_name;
 
   /* Verify if the 1st and 2nd byte in the nvlist are valid. */
   /* NOTE: independently of what endianness header announces all 
@@ -3104,62 +3234,18 @@ nvlist_find_value (const char *nvlist_in, const char *name,
       return 0;
     }
 
-  /* skip the header, nvl_version, and nvl_nvflag */
-  nvlist = nvlist + 4 * 3;
   /*
    * Loop thru the nvpair list
    * The XDR representation of an integer is in big-endian byte order.
    */
-  while ((encode_size = grub_be_to_cpu32 (grub_get_unaligned32 (nvlist))))
+  while ((nvpair=nvlist_next_nvpair(nvlist,nvpair)))
     {
-      int nelm;
-
-      if (nvlist + 4 * 4 >= nvlist_in + VDEV_PHYS_SIZE)
+      nvpair_name(nvpair,&nvp_name,&name_len);
+      type = nvpair_type(nvpair);
+      if ((grub_strncmp (nvp_name, name, grub_strlen(name)) == 0) && type == valtype)
 	{
-	  grub_dprintf ("zfs", "nvlist overflow\n");
-	  grub_error (GRUB_ERR_BAD_FS, "incorrect nvlist");
-	  return 0;
-	}
-
-      nvpair = nvlist + 4 * 2;	/* skip the encode/decode size */
-
-      name_len = grub_be_to_cpu32 (grub_get_unaligned32 (nvpair));
-      nvpair += 4;
-
-      nvp_name = nvpair;
-      nvpair = nvpair + ((name_len + 3) & ~3);	/* align */
-
-      if (nvpair + 8 >= nvlist_in + VDEV_PHYS_SIZE
-	  || encode_size < 0
-	  || nvpair + 8 + encode_size > nvlist_in + VDEV_PHYS_SIZE)
-	{
-	  grub_dprintf ("zfs", "nvlist overflow\n");
-	  grub_error (GRUB_ERR_BAD_FS, "incorrect nvlist");
-	  return 0;
+	  return nvpair_value(nvpair,val,size_out,nelm_out);
 	}
-
-      type = grub_be_to_cpu32 (grub_get_unaligned32 (nvpair));
-      nvpair += 4;
-
-      nelm = grub_be_to_cpu32 (grub_get_unaligned32 (nvpair));
-      if (nelm < 1)
-	{
-	  grub_error (GRUB_ERR_BAD_FS, "empty nvpair");
-	  return 0;
-	}
-
-      nvpair += 4;
-
-      if ((grub_strncmp (nvp_name, name, name_len) == 0) && type == valtype)
-	{
-	  *val = (char *) nvpair;
-	  *size_out = encode_size;
-	  if (nelm_out)
-	    *nelm_out = nelm;
-	  return 1;
-	}
-
-      nvlist += encode_size;	/* goto the next nvpair */
     }
   return 0;
 }


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.5: 0004-Check-for-feature-flags-needed-to-read-from-the-pool.patch --]
[-- Type: text/x-patch; name="0004-Check-for-feature-flags-needed-to-read-from-the-pool.patch", Size: 7462 bytes --]

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


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.6: 0005-LZ4-compression-in-ZFS.patch --]
[-- Type: text/x-patch; name="0005-LZ4-compression-in-ZFS.patch", Size: 12392 bytes --]

diff --git a/Makefile.util.def b/Makefile.util.def
index 3ee5e4e..3ab043a 100644
--- a/Makefile.util.def
+++ b/Makefile.util.def
@@ -96,6 +96,7 @@ library = {
   common = grub-core/fs/zfs/zfs.c;
   common = grub-core/fs/zfs/zfsinfo.c;
   common = grub-core/fs/zfs/zfs_lzjb.c;
+  common = grub-core/fs/zfs/zfs_lz4.c;
   common = grub-core/fs/zfs/zfs_sha256.c;
   common = grub-core/fs/zfs/zfs_fletcher.c;
   common = grub-core/lib/envblk.c;
diff --git a/grub-core/Makefile.core.def b/grub-core/Makefile.core.def
index 4609a4b..94dc6c9 100644
--- a/grub-core/Makefile.core.def
+++ b/grub-core/Makefile.core.def
@@ -1224,6 +1224,7 @@ module = {
   name = zfs;
   common = fs/zfs/zfs.c;
   common = fs/zfs/zfs_lzjb.c;
+  common = fs/zfs/zfs_lz4.c;
   common = fs/zfs/zfs_sha256.c;
   common = fs/zfs/zfs_fletcher.c;
 };
diff --git a/grub-core/fs/zfs/zfs.c b/grub-core/fs/zfs/zfs.c
index cf10534..aaf1af4 100644
--- a/grub-core/fs/zfs/zfs.c
+++ b/grub-core/fs/zfs/zfs.c
@@ -154,11 +154,13 @@ ZAP_LEAF_ENTRY(zap_leaf_phys_t *l, int bs, int idx)
 
 
 /*
- * Decompression Entry - lzjb
+ * Decompression Entry - lzjb & lz4
  */
 
 extern grub_err_t lzjb_decompress (void *, void *, grub_size_t, grub_size_t);
 
+extern grub_err_t lz4_decompress (void *, void *, grub_size_t, grub_size_t);
+
 typedef grub_err_t zfs_decomp_func_t (void *s_start, void *d_start,
 				      grub_size_t s_len, grub_size_t d_len);
 typedef struct decomp_entry
@@ -278,7 +280,7 @@ grub_crypto_cipher_handle_t (*grub_zfs_load_key) (const struct grub_zfs_key *key
  * to be listed here since grub opens pools in read-only mode.
  */
 static const char *spa_feature_names[] = {
-	"max.test:feat1",NULL
+	"org.illumos:lz4_compress",NULL
 };
 
 static int
@@ -344,6 +346,7 @@ static decomp_entry_t decomp_table[ZIO_COMPRESS_FUNCTIONS] = {
   {"gzip-8", zlib_decompress},  /* ZIO_COMPRESS_GZIP8 */
   {"gzip-9", zlib_decompress},  /* ZIO_COMPRESS_GZIP9 */
   {"zle", zle_decompress},      /* ZIO_COMPRESS_ZLE   */
+  {"lz4", lz4_decompress},      /* ZIO_COMPRESS_LZ4   */
 };
 
 static grub_err_t zio_read_data (blkptr_t * bp, grub_zfs_endian_t endian,
diff --git a/grub-core/fs/zfs/zfs_lz4.c b/grub-core/fs/zfs/zfs_lz4.c
new file mode 100644
index 0000000..f199434
--- /dev/null
+++ b/grub-core/fs/zfs/zfs_lz4.c
@@ -0,0 +1,321 @@
+/*
+ * LZ4 - Fast LZ compression algorithm
+ * Header File
+ * Copyright (C) 2011-2013, Yann Collet.
+ * BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * You can contact the author at :
+ * - LZ4 homepage : http://fastcompression.blogspot.com/p/lz4.html
+ * - LZ4 source repository : http://code.google.com/p/lz4/
+ */
+
+#include <grub/err.h>
+#include <grub/mm.h>
+#include <grub/misc.h>
+#include <grub/types.h>
+
+static int LZ4_uncompress_unknownOutputSize(const char *source, char *dest,
+					    int isize, int maxOutputSize);
+
+/*
+ * CPU Feature Detection
+ */
+
+/* 32 or 64 bits ? */
+#if (defined(__x86_64__) || defined(__x86_64) || defined(__amd64__) || \
+	defined(__amd64) || defined(__ppc64__) || defined(_WIN64) || \
+	defined(__LP64__) || defined(_LP64))
+#define	LZ4_ARCH64	1
+#else
+#define	LZ4_ARCH64	0
+#endif
+
+/*
+ * Little Endian or Big Endian?
+ * Note: overwrite the below #define if you know your architecture endianess.
+ */
+#if (defined(__BIG_ENDIAN__) || defined(__BIG_ENDIAN) || \
+	defined(_BIG_ENDIAN) || defined(_ARCH_PPC) || defined(__PPC__) || \
+	defined(__PPC) || defined(PPC) || defined(__powerpc__) || \
+	defined(__powerpc) || defined(powerpc) || \
+	((defined(__BYTE_ORDER__)&&(__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__))))
+#define	LZ4_BIG_ENDIAN	1
+#else
+	/*
+	 * Little Endian assumed. PDP Endian and other very rare endian format
+	 * are unsupported.
+	 */
+#endif
+
+/*
+ * Compiler Options
+ */
+
+#if __STDC_VERSION__ >= 199901L	/* C99 */
+/* "restrict" is a known keyword */
+#else
+/* Disable restrict */
+#ifndef restrict
+#define	restrict /* Only if somebody already didn't take care of that.*/
+#endif
+#endif
+
+#define	GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
+
+#define	lz4_bswap16(x) ((unsigned short int) ((((x) >> 8) & 0xffu) \
+	| (((x) & 0xffu) << 8)))
+
+#if (GCC_VERSION >= 302) || (__INTEL_COMPILER >= 800) || defined(__clang__)
+#define	expect(expr, value)    (__builtin_expect((expr), (value)))
+#else
+#define	expect(expr, value)    (expr)
+#endif
+
+#define	likely(expr)	expect((expr) != 0, 1)
+#define	unlikely(expr)	expect((expr) != 0, 0)
+
+/* Basic types */
+#define	BYTE	grub_uint8_t
+#define	U16	grub_uint16_t
+#define	U32	grub_uint32_t
+#define	S32	grub_int32_t
+#define	U64	grub_uint64_t
+typedef grub_size_t size_t;
+
+typedef struct _U16_S {
+	U16 v;
+} U16_S;
+typedef struct _U32_S {
+	U32 v;
+} U32_S;
+typedef struct _U64_S {
+	U64 v;
+} U64_S;
+
+#define	A64(x)	(((U64_S *)(x))->v)
+#define	A32(x)	(((U32_S *)(x))->v)
+#define	A16(x)	(((U16_S *)(x))->v)
+
+/*
+ * Constants
+ */
+#define	MINMATCH 4
+
+#define	COPYLENGTH 8
+#define	LASTLITERALS 5
+
+#define	ML_BITS 4
+#define	ML_MASK ((1U<<ML_BITS)-1)
+#define	RUN_BITS (8-ML_BITS)
+#define	RUN_MASK ((1U<<RUN_BITS)-1)
+
+/*
+ * Architecture-specific macros
+ */
+#if LZ4_ARCH64
+#define	STEPSIZE 8
+#define	UARCH U64
+#define	AARCH A64
+#define	LZ4_COPYSTEP(s, d)	A64(d) = A64(s); d += 8; s += 8;
+#define	LZ4_COPYPACKET(s, d)	LZ4_COPYSTEP(s, d)
+#define	LZ4_SECURECOPY(s, d, e)	if (d < e) LZ4_WILDCOPY(s, d, e)
+#define	HTYPE U32
+#define	INITBASE(base)		const BYTE* const base = ip
+#else
+#define	STEPSIZE 4
+#define	UARCH U32
+#define	AARCH A32
+#define	LZ4_COPYSTEP(s, d)	A32(d) = A32(s); d += 4; s += 4;
+#define	LZ4_COPYPACKET(s, d)	LZ4_COPYSTEP(s, d); LZ4_COPYSTEP(s, d);
+#define	LZ4_SECURECOPY		LZ4_WILDCOPY
+#define	HTYPE const BYTE*
+#define	INITBASE(base)		const int base = 0
+#endif
+
+#if (defined(LZ4_BIG_ENDIAN) && !defined(BIG_ENDIAN_NATIVE_BUT_INCOMPATIBLE))
+#define	LZ4_READ_LITTLEENDIAN_16(d, s, p) \
+	{ U16 v = A16(p); v = lz4_bswap16(v); d = (s) - v; }
+#define	LZ4_WRITE_LITTLEENDIAN_16(p, i) \
+	{ U16 v = (U16)(i); v = lz4_bswap16(v); A16(p) = v; p += 2; }
+#else
+#define	LZ4_READ_LITTLEENDIAN_16(d, s, p) { d = (s) - A16(p); }
+#define	LZ4_WRITE_LITTLEENDIAN_16(p, v)  { A16(p) = v; p += 2; }
+#endif
+
+/* Macros */
+#define	LZ4_WILDCOPY(s, d, e) do { LZ4_COPYPACKET(s, d) } while (d < e);
+
+/* Decompression functions */
+grub_err_t
+lz4_decompress(void *s_start, void *d_start, size_t s_len, size_t d_len);
+
+grub_err_t
+lz4_decompress(void *s_start, void *d_start, size_t s_len, size_t d_len)
+{
+	const BYTE *src = s_start;
+	U32 bufsiz = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) |
+	    src[3];
+
+	/* invalid compressed buffer size encoded at start */
+	if (bufsiz + 4 > s_len)
+		return grub_error(GRUB_ERR_BAD_FS,"lz4 decompression failed.");
+
+	/*
+	 * Returns 0 on success (decompression function returned non-negative)
+	 * and appropriate error on failure (decompression function returned negative).
+	 */
+	return (LZ4_uncompress_unknownOutputSize((char*)s_start + 4, d_start, bufsiz,
+	    d_len) < 0)?grub_error(GRUB_ERR_BAD_FS,"lz4 decompression failed."):0;
+}
+
+static int
+LZ4_uncompress_unknownOutputSize(const char *source,
+    char *dest, int isize, int maxOutputSize)
+{
+	/* Local Variables */
+	const BYTE *restrict ip = (const BYTE *) source;
+	const BYTE *const iend = ip + isize;
+	const BYTE *restrict ref;
+
+	BYTE *restrict op = (BYTE *) dest;
+	BYTE *const oend = op + maxOutputSize;
+	BYTE *cpy;
+
+	size_t dec[] = { 0, 3, 2, 3, 0, 0, 0, 0 };
+
+	/* Main Loop */
+	while (ip < iend) {
+		BYTE token;
+		int length;
+
+		/* get runlength */
+		token = *ip++;
+		if ((length = (token >> ML_BITS)) == RUN_MASK) {
+			int s = 255;
+			while ((ip < iend) && (s == 255)) {
+				s = *ip++;
+				length += s;
+			}
+		}
+		/* copy literals */
+		cpy = op + length;
+		if ((cpy > oend - COPYLENGTH) ||
+		    (ip + length > iend - COPYLENGTH)) {
+			if (cpy > oend)
+				/*
+				 * Error: request to write beyond destination
+				 * buffer.
+				 */
+				goto _output_error;
+			if (ip + length > iend)
+				/*
+				 * Error : request to read beyond source
+				 * buffer.
+				 */
+				goto _output_error;
+			grub_memcpy(op, ip, length);
+			op += length;
+			ip += length;
+			if (ip < iend)
+				/* Error : LZ4 format violation */
+				goto _output_error;
+			/* Necessarily EOF, due to parsing restrictions. */
+			break;
+		}
+		LZ4_WILDCOPY(ip, op, cpy);
+		ip -= (op - cpy);
+		op = cpy;
+
+		/* get offset */
+		LZ4_READ_LITTLEENDIAN_16(ref, cpy, ip);
+		ip += 2;
+		if (ref < (BYTE * const) dest)
+			/*
+			 * Error: offset creates reference outside of
+			 * destination buffer.
+			 */
+			goto _output_error;
+
+		/* get matchlength */
+		if ((length = (token & ML_MASK)) == ML_MASK) {
+			while (ip < iend) {
+				int s = *ip++;
+				length += s;
+				if (s == 255)
+					continue;
+				break;
+			}
+		}
+		/* copy repeated sequence */
+		if unlikely(op - ref < STEPSIZE) {
+#if LZ4_ARCH64
+			size_t dec2table[] = { 0, 0, 0, -1, 0, 1, 2, 3 };
+			size_t dec2 = dec2table[op - ref];
+#else
+			const int dec2 = 0;
+#endif
+			*op++ = *ref++;
+			*op++ = *ref++;
+			*op++ = *ref++;
+			*op++ = *ref++;
+			ref -= dec[op - ref];
+			A32(op) = A32(ref);
+			op += STEPSIZE - 4;
+			ref -= dec2;
+		} else {
+			LZ4_COPYSTEP(ref, op);
+		}
+		cpy = op + length - (STEPSIZE - 4);
+		if (cpy > oend - COPYLENGTH) {
+			if (cpy > oend)
+				/*
+				 * Error: request to write outside of
+				 * destination buffer.
+				 */
+				goto _output_error;
+			LZ4_SECURECOPY(ref, op, (oend - COPYLENGTH));
+			while (op < cpy)
+				*op++ = *ref++;
+			op = cpy;
+			if (op == oend)
+				/*
+				 * Check EOF (should never happen, since last
+				 * 5 bytes are supposed to be literals).
+				 */
+				break;
+			continue;
+		}
+		LZ4_SECURECOPY(ref, op, cpy);
+		op = cpy;	/* correction */
+	}
+
+	/* end of decoding */
+	return (int)(((char *)op) - dest);
+
+	/* write overflow error detected */
+	_output_error:
+	return (int)(-(((char *)ip) - source));
+}
diff --git a/include/grub/zfs/zio.h b/include/grub/zfs/zio.h
index b1c46da..8fad2cc 100644
--- a/include/grub/zfs/zio.h
+++ b/include/grub/zfs/zio.h
@@ -88,6 +88,7 @@ enum zio_compress {
 	ZIO_COMPRESS_GZIP8,
 	ZIO_COMPRESS_GZIP9,
 	ZIO_COMPRESS_ZLE,
+	ZIO_COMPRESS_LZ4,
 	ZIO_COMPRESS_FUNCTIONS
 };
 
diff --git a/po/POTFILES.in b/po/POTFILES.in
index 01cc53c..a0a5d60 100644
--- a/po/POTFILES.in
+++ b/po/POTFILES.in
@@ -174,6 +174,7 @@
 ./grub-core/fs/zfs/zfs_fletcher.c
 ./grub-core/fs/zfs/zfsinfo.c
 ./grub-core/fs/zfs/zfs_lzjb.c
+./grub-core/fs/zfs/zfs_lz4.c
 ./grub-core/fs/zfs/zfs_sha256.c
 ./grub-core/gdb/cstub.c
 ./grub-core/gdb/gdb.c


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

Re: [PATCH 1/5] Fix reporting of RAIDZ vdevs in zfsinfo

[Massimo Maggi]

[-- Attachment #1.1: Type: text/plain, Size: 728 bytes --]

On 06/05/2013 23:03, Massimo Maggi wrote:
> 
> When the issue about the infinite loop gets fully understood and fixed,
> I'll send an updated set of patches.

Sorry for the long wait, I was really busy.
I've seen that the previous patches were merged. :)
In https://github.com/maxximino/grub2/issues/1,
Christopher Siden suggested as a proper fix to check the zio_eck_t
structure inside the vdev_phys_t structure, which includes a magic
number, to be sure that the code is looking as a real ZFS filesystem.
I've added also the validation of the SHA256 checksum of the structure
itself, so we know that we are not dealing with corrupted or improper data.
You can find the patch attached.
Regards,
Massimo Maggi


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.2: 0001-ZFS-label-detection-improvements.patch --]
[-- Type: text/x-patch; name="0001-ZFS-label-detection-improvements.patch", Size: 2225 bytes --]

From 82eb1143c7bfef816d3f875041c5d65e7a69c8a5 Mon Sep 17 00:00:00 2001
From: Massimo Maggi <me@massimo-maggi.eu>
Date: Fri, 9 Aug 2013 00:39:33 +0200
Subject: [PATCH] ZFS label detection improvements.

In order to not let zfs code mistakenly analyze a block device as a ZFS
filesystem when it contains some other filesystem,let's analyze better
the label and fail as soon as possible.  In details:

     * Check the magic number of the zio_eck_t in the vdev_phys_t structure,
       which is the first possible magic number met when analyzing a block
       device.
     * Verify the SHA256 checksum of the vdev_phys_t structure, which
       contains the nvlists that are going to be read later.

Signed-off-by: Massimo Maggi<me@massimo-maggi.eu>
---
 grub-core/fs/zfs/zfs.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/grub-core/fs/zfs/zfs.c b/grub-core/fs/zfs/zfs.c
index f4acfd8..ceb4df7 100644
--- a/grub-core/fs/zfs/zfs.c
+++ b/grub-core/fs/zfs/zfs.c
@@ -969,6 +969,9 @@ check_pool_label (struct grub_zfs_data *data,
   grub_uint64_t version;
   int found;
   grub_err_t err;
+  int endian;
+  vdev_phys_t *phys;
+  zio_cksum_t emptycksum;
 
   *inserted = 0;
 
@@ -976,6 +979,28 @@ check_pool_label (struct grub_zfs_data *data,
   if (err)
     return err;
 
+  phys=(vdev_phys_t*)nvlist;
+  if( grub_zfs_to_cpu64(phys->vp_zbt.zec_magic,
+	  GRUB_ZFS_LITTLE_ENDIAN) == ZEC_MAGIC)
+    {
+      endian = GRUB_ZFS_LITTLE_ENDIAN;
+    }
+  else if(grub_zfs_to_cpu64(phys->vp_zbt.zec_magic,
+	  GRUB_ZFS_BIG_ENDIAN) != ZEC_MAGIC)
+    {
+      endian = GRUB_ZFS_BIG_ENDIAN;
+    }
+  else
+    {
+      grub_error (GRUB_ERR_BAD_FS, "bad vdev_phys_t.vp_zbt.zec_magic number");
+      return grub_errno;
+    }
+  // Now check the integrity of the vdev_phys_t structure though checksum.
+  ZIO_SET_CHECKSUM(&emptycksum, diskdesc->vdev_phys_sector << 9, 0, 0, 0);
+  err = zio_checksum_verify (emptycksum, ZIO_CHECKSUM_LABEL, endian, nvlist, VDEV_PHYS_SIZE);
+  if(err)
+     return err;
+
   grub_dprintf ("zfs", "check 2 passed\n");
 
   found = grub_zfs_nvlist_lookup_uint64 (nvlist, ZPOOL_CONFIG_POOL_STATE,
-- 
1.8.2.1


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

Re: [PATCH 1/5] Fix reporting of RAIDZ vdevs in zfsinfo

[Andrey Borzenkov]

[-- Attachment #1: Type: text/plain, Size: 1272 bytes --]

В Fri, 09 Aug 2013 02:00:51 +0200
Massimo Maggi <me@massimo-maggi.eu> пишет:

> On 06/05/2013 23:03, Massimo Maggi wrote:
> > 
> > When the issue about the infinite loop gets fully understood and fixed,
> > I'll send an updated set of patches.
> 
> Sorry for the long wait, I was really busy.
> I've seen that the previous patches were merged. :)
> In https://github.com/maxximino/grub2/issues/1,
> Christopher Siden suggested as a proper fix to check the zio_eck_t
> structure inside the vdev_phys_t structure, which includes a magic
> number, to be sure that the code is looking as a real ZFS filesystem.
> I've added also the validation of the SHA256 checksum of the structure
> itself, so we know that we are not dealing with corrupted or improper data.
> You can find the patch attached.

> +  phys=(vdev_phys_t*)nvlist;
> +  if( grub_zfs_to_cpu64(phys->vp_zbt.zec_magic,
> +	  GRUB_ZFS_LITTLE_ENDIAN) == ZEC_MAGIC)
> +    {
> +      endian = GRUB_ZFS_LITTLE_ENDIAN;
> +    }
> +  else if(grub_zfs_to_cpu64(phys->vp_zbt.zec_magic,
> +	  GRUB_ZFS_BIG_ENDIAN) != ZEC_MAGIC)

This condition sounds strange. Should it not be " == ZEC_MAGIC"?

> +    {
> +      endian = GRUB_ZFS_BIG_ENDIAN;
> +    }


> Regards,
> Massimo Maggi
> 


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [PATCH 1/5] Fix reporting of RAIDZ vdevs in zfsinfo

[Massimo Maggi]

[-- Attachment #1.1: Type: text/plain, Size: 262 bytes --]

Nella citazione in data ven 09 ago 2013 04:39:35 CEST, Andrey Borzenkov
ha scritto:
>
> This condition sounds strange. Should it not be " == ZEC_MAGIC"?
>

Obviously you're right! Thanks for catching this!
I've attached the updated patch.
Massimo Maggi

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.2: 0001-ZFS-label-detection-improvements.patch --]
[-- Type: text/x-patch; name="0001-ZFS-label-detection-improvements.patch", Size: 2225 bytes --]

From 7270d3fe49fbaf168897f84b0cc75cd4450264f6 Mon Sep 17 00:00:00 2001
From: Massimo Maggi <me@massimo-maggi.eu>
Date: Fri, 9 Aug 2013 00:39:33 +0200
Subject: [PATCH] ZFS label detection improvements.

In order to not let zfs code mistakenly analyze a block device as a ZFS
filesystem when it contains some other filesystem,let's analyze better
the label and fail as soon as possible.  In details:

     * Check the magic number of the zio_eck_t in the vdev_phys_t structure,
       which is the first possible magic number met when analyzing a block
       device.
     * Verify the SHA256 checksum of the vdev_phys_t structure, which
       contains the nvlists that are going to be read later.

Signed-off-by: Massimo Maggi<me@massimo-maggi.eu>
---
 grub-core/fs/zfs/zfs.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/grub-core/fs/zfs/zfs.c b/grub-core/fs/zfs/zfs.c
index f4acfd8..54d2612 100644
--- a/grub-core/fs/zfs/zfs.c
+++ b/grub-core/fs/zfs/zfs.c
@@ -969,6 +969,9 @@ check_pool_label (struct grub_zfs_data *data,
   grub_uint64_t version;
   int found;
   grub_err_t err;
+  int endian;
+  vdev_phys_t *phys;
+  zio_cksum_t emptycksum;
 
   *inserted = 0;
 
@@ -976,6 +979,28 @@ check_pool_label (struct grub_zfs_data *data,
   if (err)
     return err;
 
+  phys=(vdev_phys_t*)nvlist;
+  if( grub_zfs_to_cpu64(phys->vp_zbt.zec_magic,
+	  GRUB_ZFS_LITTLE_ENDIAN) == ZEC_MAGIC)
+    {
+      endian = GRUB_ZFS_LITTLE_ENDIAN;
+    }
+  else if(grub_zfs_to_cpu64(phys->vp_zbt.zec_magic,
+	  GRUB_ZFS_BIG_ENDIAN) == ZEC_MAGIC)
+    {
+      endian = GRUB_ZFS_BIG_ENDIAN;
+    }
+  else
+    {
+      grub_error (GRUB_ERR_BAD_FS, "bad vdev_phys_t.vp_zbt.zec_magic number");
+      return grub_errno;
+    }
+  // Now check the integrity of the vdev_phys_t structure though checksum.
+  ZIO_SET_CHECKSUM(&emptycksum, diskdesc->vdev_phys_sector << 9, 0, 0, 0);
+  err = zio_checksum_verify (emptycksum, ZIO_CHECKSUM_LABEL, endian, nvlist, VDEV_PHYS_SIZE);
+  if(err)
+     return err;
+
   grub_dprintf ("zfs", "check 2 passed\n");
 
   found = grub_zfs_nvlist_lookup_uint64 (nvlist, ZPOOL_CONFIG_POOL_STATE,
-- 
1.8.2.1


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

[PATCH 2/5] Accept pool version 5000 for feature flags

[Massimo Maggi]

[-- Attachment #1.1: Type: text/plain, Size: 2664 bytes --]

Ported to GRUB2 from IllumOS changeset 13700	2889e2596bd6

Author: Christopher Siden <chris.siden@delphix.com>
Ported-by: Massimo Maggi <me@massimo-maggi.eu>
---
 grub-core/fs/zfs/zfs.c | 14 +++++---------
 include/grub/zfs/zfs.h |  9 +++++++--
 2 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/grub-core/fs/zfs/zfs.c b/grub-core/fs/zfs/zfs.c
index 822d65b..3ea9c95 100644
--- a/grub-core/fs/zfs/zfs.c
+++ b/grub-core/fs/zfs/zfs.c
@@ -490,15 +490,11 @@ uberblock_verify (uberblock_phys_t * ub,
grub_uint64_t offset,
    if (grub_zfs_to_cpu64 (uber->ub_magic, GRUB_ZFS_LITTLE_ENDIAN)
       == UBERBLOCK_MAGIC
-      && grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_LITTLE_ENDIAN) >
0 -      && grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_LITTLE_ENDIAN)
-      <= SPA_VERSION)
-    endian = GRUB_ZFS_LITTLE_ENDIAN;
+      && SPA_VERSION_IS_SUPPORTED(grub_zfs_to_cpu64 (uber->ub_version,
GRUB_ZFS_LITTLE_ENDIAN)))
+     endian = GRUB_ZFS_LITTLE_ENDIAN;
    if (grub_zfs_to_cpu64 (uber->ub_magic, GRUB_ZFS_BIG_ENDIAN) ==
UBERBLOCK_MAGIC
-      && grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_BIG_ENDIAN) > 0
-      && grub_zfs_to_cpu64 (uber->ub_version, GRUB_ZFS_BIG_ENDIAN)
-      <= SPA_VERSION)
+      && SPA_VERSION_IS_SUPPORTED(grub_zfs_to_cpu64 (uber->ub_version,
GRUB_ZFS_BIG_ENDIAN)))
     endian = GRUB_ZFS_BIG_ENDIAN;
    if (endian == GRUB_ZFS_UNKNOWN_ENDIAN)
@@ -845,13 +841,13 @@ check_pool_label (struct grub_zfs_data *data,
     }
   grub_dprintf ("zfs", "check 8 passed\n");
 -  if (version > SPA_VERSION)
+  if (!SPA_VERSION_IS_SUPPORTED(version))
     {
       grub_free (nvlist);
       return grub_error (GRUB_ERR_NOT_IMPLEMENTED_YET,
 			 "too new version %llu > %llu",
 			 (unsigned long long) version,
-			 (unsigned long long) SPA_VERSION);
+			 (unsigned long long) SPA_VERSION_BEFORE_FEATURES);
     }
   grub_dprintf ("zfs", "check 9 passed\n");
 diff --git a/include/grub/zfs/zfs.h b/include/grub/zfs/zfs.h
index e326c8b..20b23b7 100644
--- a/include/grub/zfs/zfs.h
+++ b/include/grub/zfs/zfs.h
@@ -36,8 +36,13 @@ typedef enum grub_zfs_endian
 /*
  * On-disk version number.
  */
-#define	SPA_VERSION			33ULL
-
+#define	SPA_VERSION_INITIAL		1ULL
+#define	SPA_VERSION_BEFORE_FEATURES	33ULL
+#define	SPA_VERSION			5000ULL
+#define	SPA_VERSION_FEATURES		5000ULL
+#define	SPA_VERSION_IS_SUPPORTED(v) \
+	(((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \
+	((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))
 /*
  * The following are configuration names used in the nvlist describing
a pool's
  * configuration.
-- 
1.8.1.1



[-- Attachment #1.2: Attached Message Part --]
[-- Type: text/plain, Size: 0 bytes --]



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

[PATCH 3/5] Refactoring of function nvlist_find_value

[Massimo Maggi]

[-- Attachment #1.1: Type: text/plain, Size: 8455 bytes --]

It's needed to allow iteration of nvlists in other functions.

Ported to GRUB2 from IllumOS changeset 13700    2889e2596bd6

Author: Christopher Siden <chris.siden@delphix.com>
Ported-by: Massimo Maggi <me@massimo-maggi.eu>
---
 grub-core/fs/zfs/zfs.c | 230
+++++++++++++++++++++++++++++++++----------------
 1 file changed, 158 insertions(+), 72 deletions(-)

diff --git a/grub-core/fs/zfs/zfs.c b/grub-core/fs/zfs/zfs.c
index 3ea9c95..7f76445 100644
--- a/grub-core/fs/zfs/zfs.c
+++ b/grub-core/fs/zfs/zfs.c
@@ -2,6 +2,7 @@
  *  GRUB  --  GRand Unified Bootloader
  *  Copyright (C) 1999,2000,2001,2002,2003,2004,2009,2010,2011  Free
Software Foundation, Inc.
  *  Copyright 2010  Sun Microsystems, Inc.
+ *  Copyright (c) 2012 by Delphix. All rights reserved.
  *
  *  GRUB is free software; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
@@ -768,6 +769,155 @@ fill_vdev_info (struct grub_zfs_data *data,
 }
  /*
+ * For a given XDR packed nvlist, verify the first 4 bytes and move on.
+ *
+ * An XDR packed nvlist is encoded as (comments from nvs_xdr_create) :
+ *
+ *      encoding method/host endian     (4 bytes)
+ *      nvl_version                     (4 bytes)
+ *      nvl_nvflag                      (4 bytes)
+ *	encoded nvpairs:
+ *		encoded size of the nvpair      (4 bytes)
+ *		decoded size of the nvpair      (4 bytes)
+ *		name string size                (4 bytes)
+ *		name string data                (sizeof(NV_ALIGN4(string))
+ *		data type                       (4 bytes)
+ *		# of elements in the nvpair     (4 bytes)
+ *		data
+ *      2 zero's for the last nvpair
+ *		(end of the entire list)	(8 bytes)
+ *
+ */
+
+/*
+ * The nvlist_next_nvpair() function returns a handle to the next
nvpair in the
+ * list following nvpair. If nvpair is NULL, the first pair is returned. If
+ * nvpair is the last pair in the nvlist, NULL is returned.
+ */
+static const char *
+nvlist_next_nvpair(const char *nvl, const char *nvpair)
+{
+	const char *nvp;
+	int encode_size;
+	int name_len;
+	if (nvl == NULL)
+		return (NULL);
+
+	if (nvpair == NULL) {
+		/* skip over header, nvl_version and nvl_nvflag */
+		nvpair = nvl + 4 * 3;
+	} else {
+		/* skip to the next nvpair */
+		encode_size = grub_be_to_cpu32 (grub_get_unaligned32(nvpair));
+		nvpair += encode_size;
+	}
+	/* 8 bytes of 0 marks the end of the list */
+	if (*(grub_uint64_t*)nvpair == 0)
+		return (NULL);
+	/*consistency checks*/
+	if (nvpair + 4 * 3 >= nvl + VDEV_PHYS_SIZE)
+	{
+	  grub_dprintf ("zfs", "nvlist overflow\n");
+	  grub_error (GRUB_ERR_BAD_FS, "incorrect nvlist");
+	  return (NULL);
+	}
+	encode_size = grub_be_to_cpu32 (grub_get_unaligned32(nvpair));
+
+	nvp = nvpair + 4*2;
+	name_len = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp += 4;
+
+	nvp = nvp + ((name_len + 3) & ~3); // align +	if (nvp + 4 >= nvl +
VDEV_PHYS_SIZE                        +	    || encode_size < 0
+	    || nvp + 4 + encode_size > nvl + VDEV_PHYS_SIZE)       +	{
+	  grub_dprintf ("zfs", "nvlist overflow\n");
+	  grub_error (GRUB_ERR_BAD_FS, "incorrect nvlist");
+	  return (NULL);
+	}
+	 /* end consistency checks */
+
+	return (nvpair);
+}
+/*
+ * This function returns 0 on success and 1 on failure. On success, a
string
+ * containing the name of nvpair is saved in buf.
+ */
+static int
+nvpair_name(const char *nvp, char **buf, int* buflen)
+{
+	int len;
+
+	/* skip over encode/decode size */
+	nvp += 4 * 2;
+
+	len = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp=nvp+4;
+	
+	*buf=(char*)nvp;
+	*buflen=len;
+
+	return (0);
+}
+/*
+ * This function retrieves the value of the nvpair in the form of
enumerated
+ * type data_type_t.
+ */
+static int
+nvpair_type(const char *nvp)
+{
+	int name_len, type;
+
+	/* skip over encode/decode size */
+	nvp += 4 * 2;
+
+	/* skip over name_len */
+	name_len = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp += 4;
+
+	/* skip over name */
+	nvp = nvp + ((name_len + 3) & ~3); /* align */
+
+	type = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+
+	return (type);
+}
+static int
+nvpair_value(const char *nvp,char **val,
+		   grub_size_t *size_out, grub_size_t *nelm_out)
+{
+	int name_len,nelm,encode_size;
+
+	/* skip over encode/decode size */
+	encode_size = grub_be_to_cpu32 (grub_get_unaligned32(nvp));
+	nvp += 8;
+
+	/* skip over name_len */
+	name_len = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp += 4;
+
+	/* skip over name */
+	nvp = nvp + ((name_len + 3) & ~3); /* align */
+	
+	/* skip over type */
+	nvp += 4;
+	nelm = grub_be_to_cpu32 (grub_get_unaligned32 (nvp));
+	nvp +=4;
+	if (nelm < 1)
+	{
+	  grub_error (GRUB_ERR_BAD_FS, "empty nvpair");
+	  return 0;
+	}
+	  *val = (char *) nvp;
+	  *size_out = encode_size;
+	  if (nelm_out)
+	    *nelm_out = nelm;
+	    +	return 1;
+}
+
+/*
  * Check the disk label information and retrieve needed vdev name-value
pairs.
  *
  */
@@ -3064,34 +3214,14 @@ dnode_get_fullpath (const char *fullpath, struct
subvolume *subvol,
   return err;
 }
 -/*
- * For a given XDR packed nvlist, verify the first 4 bytes and move on.
- *
- * An XDR packed nvlist is encoded as (comments from nvs_xdr_create) :
- *
- *      encoding method/host endian     (4 bytes)
- *      nvl_version                     (4 bytes)
- *      nvl_nvflag                      (4 bytes)
- *	encoded nvpairs:
- *		encoded size of the nvpair      (4 bytes)
- *		decoded size of the nvpair      (4 bytes)
- *		name string size                (4 bytes)
- *		name string data                (sizeof(NV_ALIGN4(string))
- *		data type                       (4 bytes)
- *		# of elements in the nvpair     (4 bytes)
- *		data
- *      2 zero's for the last nvpair
- *		(end of the entire list)	(8 bytes)
- *
- */
-
 static int
 nvlist_find_value (const char *nvlist_in, const char *name,
 		   int valtype, char **val,
 		   grub_size_t *size_out, grub_size_t *nelm_out)
 {
-  int name_len, type, encode_size;
-  const char *nvpair, *nvp_name, *nvlist = nvlist_in;
+  int name_len, type ;
+  const char *nvpair=NULL,*nvlist=nvlist_in;
+  char *nvp_name;
    /* Verify if the 1st and 2nd byte in the nvlist are valid. */
   /* NOTE: independently of what endianness header announces all @@
-3104,62 +3234,18 @@ nvlist_find_value (const char *nvlist_in, const
char *name,
       return 0;
     }
 -  /* skip the header, nvl_version, and nvl_nvflag */
-  nvlist = nvlist + 4 * 3;
   /*
    * Loop thru the nvpair list
    * The XDR representation of an integer is in big-endian byte order.
    */
-  while ((encode_size = grub_be_to_cpu32 (grub_get_unaligned32 (nvlist))))
+  while ((nvpair=nvlist_next_nvpair(nvlist,nvpair)))
     {
-      int nelm;
-
-      if (nvlist + 4 * 4 >= nvlist_in + VDEV_PHYS_SIZE)
+      nvpair_name(nvpair,&nvp_name,&name_len);
+      type = nvpair_type(nvpair);
+      if ((grub_strncmp (nvp_name, name, grub_strlen(name)) == 0) &&
type == valtype)
 	{
-	  grub_dprintf ("zfs", "nvlist overflow\n");
-	  grub_error (GRUB_ERR_BAD_FS, "incorrect nvlist");
-	  return 0;
-	}
-
-      nvpair = nvlist + 4 * 2;	/* skip the encode/decode size */
-
-      name_len = grub_be_to_cpu32 (grub_get_unaligned32 (nvpair));
-      nvpair += 4;
-
-      nvp_name = nvpair;
-      nvpair = nvpair + ((name_len + 3) & ~3);	/* align */
-
-      if (nvpair + 8 >= nvlist_in + VDEV_PHYS_SIZE
-	  || encode_size < 0
-	  || nvpair + 8 + encode_size > nvlist_in + VDEV_PHYS_SIZE)
-	{
-	  grub_dprintf ("zfs", "nvlist overflow\n");
-	  grub_error (GRUB_ERR_BAD_FS, "incorrect nvlist");
-	  return 0;
+	  return nvpair_value(nvpair,val,size_out,nelm_out);
 	}
-
-      type = grub_be_to_cpu32 (grub_get_unaligned32 (nvpair));
-      nvpair += 4;
-
-      nelm = grub_be_to_cpu32 (grub_get_unaligned32 (nvpair));
-      if (nelm < 1)
-	{
-	  grub_error (GRUB_ERR_BAD_FS, "empty nvpair");
-	  return 0;
-	}
-
-      nvpair += 4;
-
-      if ((grub_strncmp (nvp_name, name, name_len) == 0) && type ==
valtype)
-	{
-	  *val = (char *) nvpair;
-	  *size_out = encode_size;
-	  if (nelm_out)
-	    *nelm_out = nelm;
-	  return 1;
-	}
-
-      nvlist += encode_size;	/* goto the next nvpair */
     }
   return 0;
 }
-- 
1.8.1.1



[-- Attachment #1.2: Attached Message Part --]
[-- Type: text/plain, Size: 0 bytes --]



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

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

[Massimo Maggi]

[-- Attachment #1.1: Type: text/plain, Size: 7859 bytes --]

Ported to GRUB2 from IllumOS changeset 13700    2889e2596bd6

Author: Christopher Siden <chris.siden@delphix.com>
Ported-by: Massimo Maggi <me@massimo-maggi.eu>
---
 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



[-- Attachment #1.2: Attached Message Part --]
[-- Type: text/plain, Size: 0 bytes --]



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

[PATCH 5/5] LZ4 compression in ZFS

[Massimo Maggi]

[-- Attachment #1.1: Type: text/plain, Size: 12959 bytes --]

IllumOS issue #3035 - original commit
a6f561b4aee75d0d028e7b36b151c8ed8a86bc76

Author:  Saso Kiselkov <skiselkov@gmail.com>
Ported-by: Massimo Maggi <me@massimo-maggi.eu>
---
 Makefile.util.def           |   1 +
 grub-core/Makefile.core.def |   1 +
 grub-core/fs/zfs/zfs.c      |   7 +-
 grub-core/fs/zfs/zfs_lz4.c  | 321
++++++++++++++++++++++++++++++++++++++++++++
 include/grub/zfs/zio.h      |   1 +
 po/POTFILES.in              |   1 +
 6 files changed, 330 insertions(+), 2 deletions(-)
 create mode 100644 grub-core/fs/zfs/zfs_lz4.c

diff --git a/Makefile.util.def b/Makefile.util.def
index 3ee5e4e..3ab043a 100644
--- a/Makefile.util.def
+++ b/Makefile.util.def
@@ -96,6 +96,7 @@ library = {
   common = grub-core/fs/zfs/zfs.c;
   common = grub-core/fs/zfs/zfsinfo.c;
   common = grub-core/fs/zfs/zfs_lzjb.c;
+  common = grub-core/fs/zfs/zfs_lz4.c;
   common = grub-core/fs/zfs/zfs_sha256.c;
   common = grub-core/fs/zfs/zfs_fletcher.c;
   common = grub-core/lib/envblk.c;
diff --git a/grub-core/Makefile.core.def b/grub-core/Makefile.core.def
index 4609a4b..94dc6c9 100644
--- a/grub-core/Makefile.core.def
+++ b/grub-core/Makefile.core.def
@@ -1224,6 +1224,7 @@ module = {
   name = zfs;
   common = fs/zfs/zfs.c;
   common = fs/zfs/zfs_lzjb.c;
+  common = fs/zfs/zfs_lz4.c;
   common = fs/zfs/zfs_sha256.c;
   common = fs/zfs/zfs_fletcher.c;
 };
diff --git a/grub-core/fs/zfs/zfs.c b/grub-core/fs/zfs/zfs.c
index cf10534..aaf1af4 100644
--- a/grub-core/fs/zfs/zfs.c
+++ b/grub-core/fs/zfs/zfs.c
@@ -154,11 +154,13 @@ ZAP_LEAF_ENTRY(zap_leaf_phys_t *l, int bs, int idx)
   /*
- * Decompression Entry - lzjb
+ * Decompression Entry - lzjb & lz4
  */
  extern grub_err_t lzjb_decompress (void *, void *, grub_size_t,
grub_size_t);
 +extern grub_err_t lz4_decompress (void *, void *, grub_size_t,
grub_size_t);
+
 typedef grub_err_t zfs_decomp_func_t (void *s_start, void *d_start,
 				      grub_size_t s_len, grub_size_t d_len);
 typedef struct decomp_entry
@@ -278,7 +280,7 @@ grub_crypto_cipher_handle_t (*grub_zfs_load_key)
(const struct grub_zfs_key *key
  * to be listed here since grub opens pools in read-only mode.
  */
 static const char *spa_feature_names[] = {
-	"max.test:feat1",NULL
+	"org.illumos:lz4_compress",NULL
 };
  static int
@@ -344,6 +346,7 @@ static decomp_entry_t
decomp_table[ZIO_COMPRESS_FUNCTIONS] = {
   {"gzip-8", zlib_decompress},  /* ZIO_COMPRESS_GZIP8 */
   {"gzip-9", zlib_decompress},  /* ZIO_COMPRESS_GZIP9 */
   {"zle", zle_decompress},      /* ZIO_COMPRESS_ZLE   */
+  {"lz4", lz4_decompress},      /* ZIO_COMPRESS_LZ4   */
 };
  static grub_err_t zio_read_data (blkptr_t * bp, grub_zfs_endian_t endian,
diff --git a/grub-core/fs/zfs/zfs_lz4.c b/grub-core/fs/zfs/zfs_lz4.c
new file mode 100644
index 0000000..f199434
--- /dev/null
+++ b/grub-core/fs/zfs/zfs_lz4.c
@@ -0,0 +1,321 @@
+/*
+ * LZ4 - Fast LZ compression algorithm
+ * Header File
+ * Copyright (C) 2011-2013, Yann Collet.
+ * BSD 2-Clause License
(http://www.opensource.org/licenses/bsd-license.php)
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * You can contact the author at :
+ * - LZ4 homepage : http://fastcompression.blogspot.com/p/lz4.html
+ * - LZ4 source repository : http://code.google.com/p/lz4/
+ */
+
+#include <grub/err.h>
+#include <grub/mm.h>
+#include <grub/misc.h>
+#include <grub/types.h>
+
+static int LZ4_uncompress_unknownOutputSize(const char *source, char *dest,
+					    int isize, int maxOutputSize);
+
+/*
+ * CPU Feature Detection
+ */
+
+/* 32 or 64 bits ? */
+#if (defined(__x86_64__) || defined(__x86_64) || defined(__amd64__) || \
+	defined(__amd64) || defined(__ppc64__) || defined(_WIN64) || \
+	defined(__LP64__) || defined(_LP64))
+#define	LZ4_ARCH64	1
+#else
+#define	LZ4_ARCH64	0
+#endif
+
+/*
+ * Little Endian or Big Endian?
+ * Note: overwrite the below #define if you know your architecture
endianess.
+ */
+#if (defined(__BIG_ENDIAN__) || defined(__BIG_ENDIAN) || \
+	defined(_BIG_ENDIAN) || defined(_ARCH_PPC) || defined(__PPC__) || \
+	defined(__PPC) || defined(PPC) || defined(__powerpc__) || \
+	defined(__powerpc) || defined(powerpc) || \
+	((defined(__BYTE_ORDER__)&&(__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__))))
+#define	LZ4_BIG_ENDIAN	1
+#else
+	/*
+	 * Little Endian assumed. PDP Endian and other very rare endian format
+	 * are unsupported.
+	 */
+#endif
+
+/*
+ * Compiler Options
+ */
+
+#if __STDC_VERSION__ >= 199901L	/* C99 */
+/* "restrict" is a known keyword */
+#else
+/* Disable restrict */
+#ifndef restrict
+#define	restrict /* Only if somebody already didn't take care of that.*/
+#endif
+#endif
+
+#define	GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
+
+#define	lz4_bswap16(x) ((unsigned short int) ((((x) >> 8) & 0xffu) \
+	| (((x) & 0xffu) << 8)))
+
+#if (GCC_VERSION >= 302) || (__INTEL_COMPILER >= 800) || defined(__clang__)
+#define	expect(expr, value)    (__builtin_expect((expr), (value)))
+#else
+#define	expect(expr, value)    (expr)
+#endif
+
+#define	likely(expr)	expect((expr) != 0, 1)
+#define	unlikely(expr)	expect((expr) != 0, 0)
+
+/* Basic types */
+#define	BYTE	grub_uint8_t
+#define	U16	grub_uint16_t
+#define	U32	grub_uint32_t
+#define	S32	grub_int32_t
+#define	U64	grub_uint64_t
+typedef grub_size_t size_t;
+
+typedef struct _U16_S {
+	U16 v;
+} U16_S;
+typedef struct _U32_S {
+	U32 v;
+} U32_S;
+typedef struct _U64_S {
+	U64 v;
+} U64_S;
+
+#define	A64(x)	(((U64_S *)(x))->v)
+#define	A32(x)	(((U32_S *)(x))->v)
+#define	A16(x)	(((U16_S *)(x))->v)
+
+/*
+ * Constants
+ */
+#define	MINMATCH 4
+
+#define	COPYLENGTH 8
+#define	LASTLITERALS 5
+
+#define	ML_BITS 4
+#define	ML_MASK ((1U<<ML_BITS)-1)
+#define	RUN_BITS (8-ML_BITS)
+#define	RUN_MASK ((1U<<RUN_BITS)-1)
+
+/*
+ * Architecture-specific macros
+ */
+#if LZ4_ARCH64
+#define	STEPSIZE 8
+#define	UARCH U64
+#define	AARCH A64
+#define	LZ4_COPYSTEP(s, d)	A64(d) = A64(s); d += 8; s += 8;
+#define	LZ4_COPYPACKET(s, d)	LZ4_COPYSTEP(s, d)
+#define	LZ4_SECURECOPY(s, d, e)	if (d < e) LZ4_WILDCOPY(s, d, e)
+#define	HTYPE U32
+#define	INITBASE(base)		const BYTE* const base = ip
+#else
+#define	STEPSIZE 4
+#define	UARCH U32
+#define	AARCH A32
+#define	LZ4_COPYSTEP(s, d)	A32(d) = A32(s); d += 4; s += 4;
+#define	LZ4_COPYPACKET(s, d)	LZ4_COPYSTEP(s, d); LZ4_COPYSTEP(s, d);
+#define	LZ4_SECURECOPY		LZ4_WILDCOPY
+#define	HTYPE const BYTE*
+#define	INITBASE(base)		const int base = 0
+#endif
+
+#if (defined(LZ4_BIG_ENDIAN) &&
!defined(BIG_ENDIAN_NATIVE_BUT_INCOMPATIBLE))
+#define	LZ4_READ_LITTLEENDIAN_16(d, s, p) \
+	{ U16 v = A16(p); v = lz4_bswap16(v); d = (s) - v; }
+#define	LZ4_WRITE_LITTLEENDIAN_16(p, i) \
+	{ U16 v = (U16)(i); v = lz4_bswap16(v); A16(p) = v; p += 2; }
+#else
+#define	LZ4_READ_LITTLEENDIAN_16(d, s, p) { d = (s) - A16(p); }
+#define	LZ4_WRITE_LITTLEENDIAN_16(p, v)  { A16(p) = v; p += 2; }
+#endif
+
+/* Macros */
+#define	LZ4_WILDCOPY(s, d, e) do { LZ4_COPYPACKET(s, d) } while (d < e);
+
+/* Decompression functions */
+grub_err_t
+lz4_decompress(void *s_start, void *d_start, size_t s_len, size_t d_len);
+
+grub_err_t
+lz4_decompress(void *s_start, void *d_start, size_t s_len, size_t d_len)
+{
+	const BYTE *src = s_start;
+	U32 bufsiz = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) |
+	    src[3];
+
+	/* invalid compressed buffer size encoded at start */
+	if (bufsiz + 4 > s_len)
+		return grub_error(GRUB_ERR_BAD_FS,"lz4 decompression failed.");
+
+	/*
+	 * Returns 0 on success (decompression function returned non-negative)
+	 * and appropriate error on failure (decompression function returned
negative).
+	 */
+	return (LZ4_uncompress_unknownOutputSize((char*)s_start + 4, d_start,
bufsiz,
+	    d_len) < 0)?grub_error(GRUB_ERR_BAD_FS,"lz4 decompression failed."):0;
+}
+
+static int
+LZ4_uncompress_unknownOutputSize(const char *source,
+    char *dest, int isize, int maxOutputSize)
+{
+	/* Local Variables */
+	const BYTE *restrict ip = (const BYTE *) source;
+	const BYTE *const iend = ip + isize;
+	const BYTE *restrict ref;
+
+	BYTE *restrict op = (BYTE *) dest;
+	BYTE *const oend = op + maxOutputSize;
+	BYTE *cpy;
+
+	size_t dec[] = { 0, 3, 2, 3, 0, 0, 0, 0 };
+
+	/* Main Loop */
+	while (ip < iend) {
+		BYTE token;
+		int length;
+
+		/* get runlength */
+		token = *ip++;
+		if ((length = (token >> ML_BITS)) == RUN_MASK) {
+			int s = 255;
+			while ((ip < iend) && (s == 255)) {
+				s = *ip++;
+				length += s;
+			}
+		}
+		/* copy literals */
+		cpy = op + length;
+		if ((cpy > oend - COPYLENGTH) ||
+		    (ip + length > iend - COPYLENGTH)) {
+			if (cpy > oend)
+				/*
+				 * Error: request to write beyond destination
+				 * buffer.
+				 */
+				goto _output_error;
+			if (ip + length > iend)
+				/*
+				 * Error : request to read beyond source
+				 * buffer.
+				 */
+				goto _output_error;
+			grub_memcpy(op, ip, length);
+			op += length;
+			ip += length;
+			if (ip < iend)
+				/* Error : LZ4 format violation */
+				goto _output_error;
+			/* Necessarily EOF, due to parsing restrictions. */
+			break;
+		}
+		LZ4_WILDCOPY(ip, op, cpy);
+		ip -= (op - cpy);
+		op = cpy;
+
+		/* get offset */
+		LZ4_READ_LITTLEENDIAN_16(ref, cpy, ip);
+		ip += 2;
+		if (ref < (BYTE * const) dest)
+			/*
+			 * Error: offset creates reference outside of
+			 * destination buffer.
+			 */
+			goto _output_error;
+
+		/* get matchlength */
+		if ((length = (token & ML_MASK)) == ML_MASK) {
+			while (ip < iend) {
+				int s = *ip++;
+				length += s;
+				if (s == 255)
+					continue;
+				break;
+			}
+		}
+		/* copy repeated sequence */
+		if unlikely(op - ref < STEPSIZE) {
+#if LZ4_ARCH64
+			size_t dec2table[] = { 0, 0, 0, -1, 0, 1, 2, 3 };
+			size_t dec2 = dec2table[op - ref];
+#else
+			const int dec2 = 0;
+#endif
+			*op++ = *ref++;
+			*op++ = *ref++;
+			*op++ = *ref++;
+			*op++ = *ref++;
+			ref -= dec[op - ref];
+			A32(op) = A32(ref);
+			op += STEPSIZE - 4;
+			ref -= dec2;
+		} else {
+			LZ4_COPYSTEP(ref, op);
+		}
+		cpy = op + length - (STEPSIZE - 4);
+		if (cpy > oend - COPYLENGTH) {
+			if (cpy > oend)
+				/*
+				 * Error: request to write outside of
+				 * destination buffer.
+				 */
+				goto _output_error;
+			LZ4_SECURECOPY(ref, op, (oend - COPYLENGTH));
+			while (op < cpy)
+				*op++ = *ref++;
+			op = cpy;
+			if (op == oend)
+				/*
+				 * Check EOF (should never happen, since last
+				 * 5 bytes are supposed to be literals).
+				 */
+				break;
+			continue;
+		}
+		LZ4_SECURECOPY(ref, op, cpy);
+		op = cpy;	/* correction */
+	}
+
+	/* end of decoding */
+	return (int)(((char *)op) - dest);
+
+	/* write overflow error detected */
+	_output_error:
+	return (int)(-(((char *)ip) - source));
+}
diff --git a/include/grub/zfs/zio.h b/include/grub/zfs/zio.h
index b1c46da..8fad2cc 100644
--- a/include/grub/zfs/zio.h
+++ b/include/grub/zfs/zio.h
@@ -88,6 +88,7 @@ enum zio_compress {
 	ZIO_COMPRESS_GZIP8,
 	ZIO_COMPRESS_GZIP9,
 	ZIO_COMPRESS_ZLE,
+	ZIO_COMPRESS_LZ4,
 	ZIO_COMPRESS_FUNCTIONS
 };
 diff --git a/po/POTFILES.in b/po/POTFILES.in
index 01cc53c..a0a5d60 100644
--- a/po/POTFILES.in
+++ b/po/POTFILES.in
@@ -174,6 +174,7 @@
 ./grub-core/fs/zfs/zfs_fletcher.c
 ./grub-core/fs/zfs/zfsinfo.c
 ./grub-core/fs/zfs/zfs_lzjb.c
+./grub-core/fs/zfs/zfs_lz4.c
 ./grub-core/fs/zfs/zfs_sha256.c
 ./grub-core/gdb/cstub.c
 ./grub-core/gdb/gdb.c
-- 
1.8.1.1


[-- Attachment #1.2: Attached Message Part --]
[-- Type: text/plain, Size: 0 bytes --]



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]