The pull request you sent on Fri, 23 Aug 2019 11:41:53 -0400:
> git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git
> tags/for-5.3/dm-fixes-2
has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/dd469a456047af5eb1ee0bcfc8fe61f5940ef0e0
Thank you!
--
In preparation for changing the path grouping code, add some unit tests
to verify that it works correctly. The only test that currently fails
(and so it being skipped) is using MULTIBUS when mp->paths is empty. All
the other path grouping policies free mp->paths, even if it is empty.
one_group() sh
In the pgpolicy functions, if an error is encountered after
alloc_pathgroup() is called, but before the path group is added to a
multipath device with add_pathgroup(), the pathgroup needs to be cleaned
up by calling free_pathgroup(). However, after the pathgroup has been
added to the multipath devi
group_paths() will now create seperate path groups for marginal and
normal paths, and place all of the marginal path groups after the normal
ones, in order by priority.
Signed-off-by: Benjamin Marzinski
---
libmultipath/pgpolicies.c | 83 +-
libmultipath/swit
The delay_checks shaky paths detection method works the same way as the
san_path_err method, but not as well, with less configurability, and
with the code spread all over check_path(). The only real difference is
that marks the path as marginal for a certain number of path checks
instead of for a s
group_paths now gets passed this to determine whether to enable
marginal pathgroups. The unit tests have also been updated.
Signed-off-by: Benjamin Marzinski
---
libmultipath/config.h | 1 +
libmultipath/configure.c | 5 +-
libmultipath/dict.c | 3 +
libmultipath/pgpolicies.c |
The pgpolicy unit tests now work again, using group_paths().
test_one_group0(), which was skipped with the old path grouping code
because it failed, is now working correctly.
Reviewed-by: Martin Wilck
Signed-off-by: Benjamin Marzinski
---
tests/pgpolicy.c | 125 +++--
group_paths() is a wrapper around the pgpolicy functions, that pulls out
the common code from the beginning and the end. For this to work,
one_group() needs to change how it sets up the pathgroups vector to work
like the other pgpolicy functions. This does means that the pathgroups
in group_by_prio
All of the code that uses vector_foreach_slot_backwards() treats "i" as
the index of the entry "p", but the way it was coded, that wasn't the
case. "i" was the number of the entry counting from 1, not 0.
Reviewed-by: Martin Wilck
Signed-off-by: Benjamin Marzinski
---
libmultipath/vector.h | 2 +
The first patch in this series is simply a resend of my previous patch
to fix vector_foreach_slot_backwards(). The rest of the patches are
related to adding support for an alternative way to deal with marginal
paths. Instead of delaying their reinstatement, which can cause problems
if multipath isn
This commit adds a marginal variable ot the paths and pathgroups structs.
The marginal paths variable can be set by
multipathd path setmarginal
and cleared by
multipathd path unsetmarginal
All of the marginal paths on a multipath device can be cleared by
multipathd map unsetmarginal
Curren
Add documentation for the marginal_pathgroups option and the
(un)setmarginal commands.
Signed-off-by: Benjamin Marzinski
---
multipath/multipath.conf.5 | 34 ++
multipathd/multipathd.8| 19 +++
2 files changed, 49 insertions(+), 4 deletions(-)
Signed-off-by: Benjamin Marzinski
---
tests/pgpolicy.c | 337 +--
1 file changed, 295 insertions(+), 42 deletions(-)
diff --git a/tests/pgpolicy.c b/tests/pgpolicy.c
index 04a77c4c..ab09f91c 100644
--- a/tests/pgpolicy.c
+++ b/tests/pgpolicy.c
@@ -40,6
To enable future changes, mp->pgpolicyfn() now takes a vector of
paths instead of always using mp->paths.
Signed-off-by: Benjamin Marzinski
---
libmultipath/pgpolicies.c | 38 +++---
libmultipath/pgpolicies.h | 10 +-
libmultipath/structs.h| 2 +-
3 f
This commit makes the marginal_pathgroups option work with the
existing methods for determining marginal paths. It also merges the
code for the marginal_path and sand_path_err methods. This has the
side effect of making the marginal_path code set a marginal path's state
to "delayed" instead of "sh
group_by_node_name() and group_by_serial() are exactly the same except
for how the paths are compared. group_by_prio() is different but its
pathvec solves the same issue as the bitmap from the other two
functions, and since we are always running sort_pathgroups() after
calling pgpriorityfn, there i
Hi Linus,
More fixes than usual from DM at this stage in a release but...
The following changes since commit d45331b00ddb179e291766617259261c112db872:
Linux 5.3-rc4 (2019-08-11 13:26:41 -0700)
are available in the Git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/device-mapp
Currently, if we pass too high sector number to dm_table_find_target, it
returns zeroed dm_target structure and callers test if the structure is
zeroed with the macro dm_target_is_valid.
However, returning NULL is common practice to indicate errors.
This patch refactors the dm code, so that dm_ta
If the sector number is too high, dm_table_find_target should return a
pointer to a zeroed dm_target structure (the caller should test it with
dm_target_is_valid).
However, for some table sizes, the code in dm_table_find_target that
performs btree lookup will access out of bound memory structures.
On Wed, Aug 21, 2019 at 17:19:41PM +0900, Herbert Xu wrote:
> On Wed, Aug 21, 2019 at 04:57:41PM +0900, boojin.kim wrote:
> >
> > Can you tell me which patch you mentioned? Is this?
> > https://patches.linaro.org/project/linux-crypto/list/?series=22762
> >
>
> Yes this is the one.
>
> Cheers,
I l
On Wed, Aug 21, 2019 at 5:10 AM Jens Axboe wrote:
>
> On 8/21/19 12:42 AM, boojin.kim wrote:
> > This patch supports crypto information to be maintained via BIO
> > and passed to the storage driver.
> >
> > To do this, 'bi_aux_private', 'REQ_CYPTE' and 'bi_dun' are added
> > to the block layer.
>
On Fri, Aug 23, 2019 at 01:28:37PM +0900, Herbert Xu wrote:
>
> No. If you're after total offload then the crypto API is not for
> you. What we can support is the offloading of encryption/decryption
> over many sectors.
>
> Cheers,
FMP doesn't use encrypt/decrypt of crypto API because it doesn't
On Wed, 22 Aug 2019 at 17:37, Krzysztof Kozlowski wrote:
>
> Your patch looks corrupted - wrapped by mailer. The easiest way
> usually is to use git format-patch and git send-email - then you do
> not have to worry about formatting etc.
>
> Best regards,
> Krzysztof
I'm using outlook instead of '
V3 just sent:
lore.kernel.org/linux-block/20190822161318.26236-1-gpicc...@canonical.com
The code was merged with the 'broken' state patch.
Cheers,
Guilherme
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
Currently md raid0/linear are not provided with any mechanism to validate
if an array member got removed or failed. The driver keeps sending BIOs
regardless of the state of array members, and kernel shows state 'clean'
in the 'array_state' sysfs attribute. This leads to the following
situation: if
> On Aug 22, 2019, at 9:13 AM, Guilherme G. Piccoli
> wrote:
>
> Currently if a md raid0/linear array gets one or more members removed while
> being mounted, kernel keeps showing state 'clean' in the 'array_state'
> sysfs attribute. Despite udev signaling the member device is gone, 'mdadm'
>
Currently if a md raid0/linear array gets one or more members removed while
being mounted, kernel keeps showing state 'clean' in the 'array_state'
sysfs attribute. Despite udev signaling the member device is gone, 'mdadm'
cannot issue the STOP_ARRAY ioctl successfully, given the array is mounted.
On 21/08/2019 16:22, Song Liu wrote:
> [...]
>
> I think this makes sense. Please send the patch and we can discuss
> further while looking at the code.
Thanks Song, just sent the V3:
lore.kernel.org/linux-block/20190822161318.26236-1-gpicc...@canonical.com
Cheers,
Guilherme
--
dm-devel mai
On 22/08/2019 05:49, Guoqing Jiang wrote:
> Hi,
Hi Guoqing, thanks for the review!
>
> On 8/16/19 3:40 PM, Guilherme G. Piccoli wrote:
>> +static bool linear_is_missing_dev(struct mddev *mddev)
>> +{
>> + struct md_rdev *rdev;
>> + static int already_missing;
>> + int def_disks, work_d
> On Aug 22, 2019, at 9:13 AM, Guilherme G. Piccoli
> wrote:
>
> Currently md raid0/linear are not provided with any mechanism to validate
> if an array member got removed or failed. The driver keeps sending BIOs
> regardless of the state of array members, and kernel shows state 'clean'
> in
Hi
I tested it and the bug is real - with some table sizes,
dm_table_find_target will access memory out of bounds if the sector
argument is beyond limit.
Your patch fixes some of these cases, but not all of them.
I used this script to test all possible table sizes:
#!/bin/bash -e
sync
dmsetup
31 matches
Mail list logo
