Re: [PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Dmitry Kasatkin]

On 04/12/17 17:40, Dmitry Kasatkin wrote:



On 04/12/17 15:42, Roberto Sassu wrote:

On 12/4/2017 1:06 PM, Mimi Zohar wrote:

Hi Dmitry,

On Fri, 2017-12-01 at 20:40 +0200, Dmitry Kasatkin wrote:
The original design was discussed 3+ years ago, but was never 
completed/upstreamed.

Based on the recent discussions with Linus
https://patchwork.kernel.org/patch/9975919, I've rebased this patch.

Before IMA appraisal was introduced, IMA was using own integrity cache
lock along with i_mutex. process_measurement and ima_file_free took
the iint->mutex first and then the i_mutex, while setxattr, chmod and
chown took the locks in reverse order. To resolve the potential 
deadlock,

i_mutex was moved to protect entire IMA functionality and the redundant
iint->mutex was eliminated.

Solution was based on the assumption that filesystem code does not take
i_mutex further. But when file is opened with O_DIRECT flag, direct-io
implementation takes i_mutex and produces deadlock. Furthermore, 
certain

other filesystem operations, such as llseek, also take i_mutex.

More recently some filesystems have replaced their filesystem specific
lock with the global i_rwsem to read a file.  As a result, when IMA
attempts to calculate the file hash, reading the file attempts to take
the i_rwsem again.

To resolve O_DIRECT related deadlock problem, this patch re-introduces
iint->mutex. But to eliminate the original chmod() related deadlock
problem, this patch eliminates the requirement for chmod hooks to take
the iint->mutex by introducing additional atomic iint->attr_flags to
indicate calling of the hooks. The allowed locking order is to take
the iint->mutex first and then the i_rwsem.

Original flags were cleared in chmod(), setxattr() or removwxattr() 
hooks
and tested when file was closed or opened again. New atomic flags 
are set
or cleared in those hooks and tested to clear iint->flags on close 
or on open.


Atomic flags are following:
* IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, 
chown, chgrp)
   and file attributes have changed. On file open, it causes IMA to 
clear

   iint->flags to re-evaluate policy and perform IMA functions again.
* IMA_CHANGE_XATTR - indicates that setxattr or removexattr was 
called and
   extended attributes have changed. On file open, it causes IMA to 
clear

   iint->flags IMA_DONE_MASK to re-appraise.
* IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
   It is cleared if file policy changes and no update is needed.
* IMA_DIGSIG - indicates that file security.ima has signature and file
   security.ima must not update to file has on file close.


Nice!  I've been testing with this patch and all seems good.  Before
queueing this patch to be upstreamed, I'd appreciate if others tested
using it as well.  It applies cleanly to the next-queued branch.


If the inode lock is released before the IMA_MEASURE flag is set, the
ToMToU violation will not be detected when a writer accesses the same
inode. This issue was fixed with commit f7a859ff7395c.

Roberto


Hi Roberto,

I will check the commit.

Dmitry


It seems you are right... That violation patch came in between locking 
patch was there. I do not remember why I have rebased it like it looks 
now. But it seems that violation checking needs to be moved under 
iint->mutex locking. Hmm. but why I have not done it like that 3 years 
ago :)


I will think how to update it.

Thanks for catching it up.

Dmitry







A subsequent patch will remove the O_DIRECT check in
ima_calc_file_hash().

Fixes: Commit 6552321831dc "xfs: remove i_iolock and use i_rwsem in
the VFS inode instead"

thanks,

Mimi

--
To unsubscribe from this list: send the line "unsubscribe 
linux-security-module" in

the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Dmitry Kasatkin]

On 04/12/17 17:40, Dmitry Kasatkin wrote:



On 04/12/17 15:42, Roberto Sassu wrote:

On 12/4/2017 1:06 PM, Mimi Zohar wrote:

Hi Dmitry,

On Fri, 2017-12-01 at 20:40 +0200, Dmitry Kasatkin wrote:
The original design was discussed 3+ years ago, but was never 
completed/upstreamed.

Based on the recent discussions with Linus
https://patchwork.kernel.org/patch/9975919, I've rebased this patch.

Before IMA appraisal was introduced, IMA was using own integrity cache
lock along with i_mutex. process_measurement and ima_file_free took
the iint->mutex first and then the i_mutex, while setxattr, chmod and
chown took the locks in reverse order. To resolve the potential 
deadlock,

i_mutex was moved to protect entire IMA functionality and the redundant
iint->mutex was eliminated.

Solution was based on the assumption that filesystem code does not take
i_mutex further. But when file is opened with O_DIRECT flag, direct-io
implementation takes i_mutex and produces deadlock. Furthermore, 
certain

other filesystem operations, such as llseek, also take i_mutex.

More recently some filesystems have replaced their filesystem specific
lock with the global i_rwsem to read a file.  As a result, when IMA
attempts to calculate the file hash, reading the file attempts to take
the i_rwsem again.

To resolve O_DIRECT related deadlock problem, this patch re-introduces
iint->mutex. But to eliminate the original chmod() related deadlock
problem, this patch eliminates the requirement for chmod hooks to take
the iint->mutex by introducing additional atomic iint->attr_flags to
indicate calling of the hooks. The allowed locking order is to take
the iint->mutex first and then the i_rwsem.

Original flags were cleared in chmod(), setxattr() or removwxattr() 
hooks
and tested when file was closed or opened again. New atomic flags 
are set
or cleared in those hooks and tested to clear iint->flags on close 
or on open.


Atomic flags are following:
* IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, 
chown, chgrp)
   and file attributes have changed. On file open, it causes IMA to 
clear

   iint->flags to re-evaluate policy and perform IMA functions again.
* IMA_CHANGE_XATTR - indicates that setxattr or removexattr was 
called and
   extended attributes have changed. On file open, it causes IMA to 
clear

   iint->flags IMA_DONE_MASK to re-appraise.
* IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
   It is cleared if file policy changes and no update is needed.
* IMA_DIGSIG - indicates that file security.ima has signature and file
   security.ima must not update to file has on file close.


Nice!  I've been testing with this patch and all seems good.  Before
queueing this patch to be upstreamed, I'd appreciate if others tested
using it as well.  It applies cleanly to the next-queued branch.


If the inode lock is released before the IMA_MEASURE flag is set, the
ToMToU violation will not be detected when a writer accesses the same
inode. This issue was fixed with commit f7a859ff7395c.

Roberto


Hi Roberto,

I will check the commit.

Dmitry


It seems you are right... That violation patch came in between locking 
patch was there. I do not remember why I have rebased it like it looks 
now. But it seems that violation checking needs to be moved under 
iint->mutex locking. Hmm. but why I have not done it like that 3 years 
ago :)


I will think how to update it.

Thanks for catching it up.

Dmitry







A subsequent patch will remove the O_DIRECT check in
ima_calc_file_hash().

Fixes: Commit 6552321831dc "xfs: remove i_iolock and use i_rwsem in
the VFS inode instead"

thanks,

Mimi

--
To unsubscribe from this list: send the line "unsubscribe 
linux-security-module" in

the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Dmitry Kasatkin]

On 04/12/17 15:42, Roberto Sassu wrote:

On 12/4/2017 1:06 PM, Mimi Zohar wrote:

Hi Dmitry,

On Fri, 2017-12-01 at 20:40 +0200, Dmitry Kasatkin wrote:
The original design was discussed 3+ years ago, but was never 
completed/upstreamed.

Based on the recent discussions with Linus
https://patchwork.kernel.org/patch/9975919, I've rebased this patch.

Before IMA appraisal was introduced, IMA was using own integrity cache
lock along with i_mutex. process_measurement and ima_file_free took
the iint->mutex first and then the i_mutex, while setxattr, chmod and
chown took the locks in reverse order. To resolve the potential 
deadlock,

i_mutex was moved to protect entire IMA functionality and the redundant
iint->mutex was eliminated.

Solution was based on the assumption that filesystem code does not take
i_mutex further. But when file is opened with O_DIRECT flag, direct-io
implementation takes i_mutex and produces deadlock. Furthermore, certain
other filesystem operations, such as llseek, also take i_mutex.

More recently some filesystems have replaced their filesystem specific
lock with the global i_rwsem to read a file.  As a result, when IMA
attempts to calculate the file hash, reading the file attempts to take
the i_rwsem again.

To resolve O_DIRECT related deadlock problem, this patch re-introduces
iint->mutex. But to eliminate the original chmod() related deadlock
problem, this patch eliminates the requirement for chmod hooks to take
the iint->mutex by introducing additional atomic iint->attr_flags to
indicate calling of the hooks. The allowed locking order is to take
the iint->mutex first and then the i_rwsem.

Original flags were cleared in chmod(), setxattr() or removwxattr() 
hooks
and tested when file was closed or opened again. New atomic flags are 
set
or cleared in those hooks and tested to clear iint->flags on close or 
on open.


Atomic flags are following:
* IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, chown, 
chgrp)
   and file attributes have changed. On file open, it causes IMA to 
clear

   iint->flags to re-evaluate policy and perform IMA functions again.
* IMA_CHANGE_XATTR - indicates that setxattr or removexattr was 
called and
   extended attributes have changed. On file open, it causes IMA to 
clear

   iint->flags IMA_DONE_MASK to re-appraise.
* IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
   It is cleared if file policy changes and no update is needed.
* IMA_DIGSIG - indicates that file security.ima has signature and file
   security.ima must not update to file has on file close.


Nice!  I've been testing with this patch and all seems good.  Before
queueing this patch to be upstreamed, I'd appreciate if others tested
using it as well.  It applies cleanly to the next-queued branch.


If the inode lock is released before the IMA_MEASURE flag is set, the
ToMToU violation will not be detected when a writer accesses the same
inode. This issue was fixed with commit f7a859ff7395c.

Roberto


Hi Roberto,

I will check the commit.

Dmitry





A subsequent patch will remove the O_DIRECT check in
ima_calc_file_hash().

Fixes: Commit 6552321831dc "xfs: remove i_iolock and use i_rwsem in
the VFS inode instead"

thanks,

Mimi

--
To unsubscribe from this list: send the line "unsubscribe 
linux-security-module" in

the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Dmitry Kasatkin]

On 04/12/17 15:42, Roberto Sassu wrote:

On 12/4/2017 1:06 PM, Mimi Zohar wrote:

Hi Dmitry,

On Fri, 2017-12-01 at 20:40 +0200, Dmitry Kasatkin wrote:
The original design was discussed 3+ years ago, but was never 
completed/upstreamed.

Based on the recent discussions with Linus
https://patchwork.kernel.org/patch/9975919, I've rebased this patch.

Before IMA appraisal was introduced, IMA was using own integrity cache
lock along with i_mutex. process_measurement and ima_file_free took
the iint->mutex first and then the i_mutex, while setxattr, chmod and
chown took the locks in reverse order. To resolve the potential 
deadlock,

i_mutex was moved to protect entire IMA functionality and the redundant
iint->mutex was eliminated.

Solution was based on the assumption that filesystem code does not take
i_mutex further. But when file is opened with O_DIRECT flag, direct-io
implementation takes i_mutex and produces deadlock. Furthermore, certain
other filesystem operations, such as llseek, also take i_mutex.

More recently some filesystems have replaced their filesystem specific
lock with the global i_rwsem to read a file.  As a result, when IMA
attempts to calculate the file hash, reading the file attempts to take
the i_rwsem again.

To resolve O_DIRECT related deadlock problem, this patch re-introduces
iint->mutex. But to eliminate the original chmod() related deadlock
problem, this patch eliminates the requirement for chmod hooks to take
the iint->mutex by introducing additional atomic iint->attr_flags to
indicate calling of the hooks. The allowed locking order is to take
the iint->mutex first and then the i_rwsem.

Original flags were cleared in chmod(), setxattr() or removwxattr() 
hooks
and tested when file was closed or opened again. New atomic flags are 
set
or cleared in those hooks and tested to clear iint->flags on close or 
on open.


Atomic flags are following:
* IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, chown, 
chgrp)
   and file attributes have changed. On file open, it causes IMA to 
clear

   iint->flags to re-evaluate policy and perform IMA functions again.
* IMA_CHANGE_XATTR - indicates that setxattr or removexattr was 
called and
   extended attributes have changed. On file open, it causes IMA to 
clear

   iint->flags IMA_DONE_MASK to re-appraise.
* IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
   It is cleared if file policy changes and no update is needed.
* IMA_DIGSIG - indicates that file security.ima has signature and file
   security.ima must not update to file has on file close.


Nice!  I've been testing with this patch and all seems good.  Before
queueing this patch to be upstreamed, I'd appreciate if others tested
using it as well.  It applies cleanly to the next-queued branch.


If the inode lock is released before the IMA_MEASURE flag is set, the
ToMToU violation will not be detected when a writer accesses the same
inode. This issue was fixed with commit f7a859ff7395c.

Roberto


Hi Roberto,

I will check the commit.

Dmitry





A subsequent patch will remove the O_DIRECT check in
ima_calc_file_hash().

Fixes: Commit 6552321831dc "xfs: remove i_iolock and use i_rwsem in
the VFS inode instead"

thanks,

Mimi

--
To unsubscribe from this list: send the line "unsubscribe 
linux-security-module" in

the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Roberto Sassu]

On 12/4/2017 1:06 PM, Mimi Zohar wrote:

Hi Dmitry,

On Fri, 2017-12-01 at 20:40 +0200, Dmitry Kasatkin wrote:

The original design was discussed 3+ years ago, but was never 
completed/upstreamed.
Based on the recent discussions with Linus
https://patchwork.kernel.org/patch/9975919, I've rebased this patch.

Before IMA appraisal was introduced, IMA was using own integrity cache
lock along with i_mutex. process_measurement and ima_file_free took
the iint->mutex first and then the i_mutex, while setxattr, chmod and
chown took the locks in reverse order. To resolve the potential deadlock,
i_mutex was moved to protect entire IMA functionality and the redundant
iint->mutex was eliminated.

Solution was based on the assumption that filesystem code does not take
i_mutex further. But when file is opened with O_DIRECT flag, direct-io
implementation takes i_mutex and produces deadlock. Furthermore, certain
other filesystem operations, such as llseek, also take i_mutex.

More recently some filesystems have replaced their filesystem specific
lock with the global i_rwsem to read a file.  As a result, when IMA
attempts to calculate the file hash, reading the file attempts to take
the i_rwsem again.

To resolve O_DIRECT related deadlock problem, this patch re-introduces
iint->mutex. But to eliminate the original chmod() related deadlock
problem, this patch eliminates the requirement for chmod hooks to take
the iint->mutex by introducing additional atomic iint->attr_flags to
indicate calling of the hooks. The allowed locking order is to take
the iint->mutex first and then the i_rwsem.

Original flags were cleared in chmod(), setxattr() or removwxattr() hooks
and tested when file was closed or opened again. New atomic flags are set
or cleared in those hooks and tested to clear iint->flags on close or on open.

Atomic flags are following:
* IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, chown, chgrp)
   and file attributes have changed. On file open, it causes IMA to clear
   iint->flags to re-evaluate policy and perform IMA functions again.
* IMA_CHANGE_XATTR - indicates that setxattr or removexattr was called and
   extended attributes have changed. On file open, it causes IMA to clear
   iint->flags IMA_DONE_MASK to re-appraise.
* IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
   It is cleared if file policy changes and no update is needed.
* IMA_DIGSIG - indicates that file security.ima has signature and file
   security.ima must not update to file has on file close.


Nice!  I've been testing with this patch and all seems good.  Before
queueing this patch to be upstreamed, I'd appreciate if others tested
using it as well.  It applies cleanly to the next-queued branch.


If the inode lock is released before the IMA_MEASURE flag is set, the
ToMToU violation will not be detected when a writer accesses the same
inode. This issue was fixed with commit f7a859ff7395c.

Roberto



A subsequent patch will remove the O_DIRECT check in
ima_calc_file_hash().

Fixes: Commit 6552321831dc "xfs: remove i_iolock and use i_rwsem in
the VFS inode instead"

thanks,

Mimi

--
To unsubscribe from this list: send the line "unsubscribe 
linux-security-module" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html



--
HUAWEI TECHNOLOGIES Duesseldorf GmbH, HRB 56063
Managing Director: Bo PENG, Qiuen PENG, Shengli WANG

Re: [PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Roberto Sassu]

On 12/4/2017 1:06 PM, Mimi Zohar wrote:

Hi Dmitry,

On Fri, 2017-12-01 at 20:40 +0200, Dmitry Kasatkin wrote:

The original design was discussed 3+ years ago, but was never 
completed/upstreamed.
Based on the recent discussions with Linus
https://patchwork.kernel.org/patch/9975919, I've rebased this patch.

Before IMA appraisal was introduced, IMA was using own integrity cache
lock along with i_mutex. process_measurement and ima_file_free took
the iint->mutex first and then the i_mutex, while setxattr, chmod and
chown took the locks in reverse order. To resolve the potential deadlock,
i_mutex was moved to protect entire IMA functionality and the redundant
iint->mutex was eliminated.

Solution was based on the assumption that filesystem code does not take
i_mutex further. But when file is opened with O_DIRECT flag, direct-io
implementation takes i_mutex and produces deadlock. Furthermore, certain
other filesystem operations, such as llseek, also take i_mutex.

More recently some filesystems have replaced their filesystem specific
lock with the global i_rwsem to read a file.  As a result, when IMA
attempts to calculate the file hash, reading the file attempts to take
the i_rwsem again.

To resolve O_DIRECT related deadlock problem, this patch re-introduces
iint->mutex. But to eliminate the original chmod() related deadlock
problem, this patch eliminates the requirement for chmod hooks to take
the iint->mutex by introducing additional atomic iint->attr_flags to
indicate calling of the hooks. The allowed locking order is to take
the iint->mutex first and then the i_rwsem.

Original flags were cleared in chmod(), setxattr() or removwxattr() hooks
and tested when file was closed or opened again. New atomic flags are set
or cleared in those hooks and tested to clear iint->flags on close or on open.

Atomic flags are following:
* IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, chown, chgrp)
   and file attributes have changed. On file open, it causes IMA to clear
   iint->flags to re-evaluate policy and perform IMA functions again.
* IMA_CHANGE_XATTR - indicates that setxattr or removexattr was called and
   extended attributes have changed. On file open, it causes IMA to clear
   iint->flags IMA_DONE_MASK to re-appraise.
* IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
   It is cleared if file policy changes and no update is needed.
* IMA_DIGSIG - indicates that file security.ima has signature and file
   security.ima must not update to file has on file close.


Nice!  I've been testing with this patch and all seems good.  Before
queueing this patch to be upstreamed, I'd appreciate if others tested
using it as well.  It applies cleanly to the next-queued branch.


If the inode lock is released before the IMA_MEASURE flag is set, the
ToMToU violation will not be detected when a writer accesses the same
inode. This issue was fixed with commit f7a859ff7395c.

Roberto



A subsequent patch will remove the O_DIRECT check in
ima_calc_file_hash().

Fixes: Commit 6552321831dc "xfs: remove i_iolock and use i_rwsem in
the VFS inode instead"

thanks,

Mimi

--
To unsubscribe from this list: send the line "unsubscribe 
linux-security-module" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html



--
HUAWEI TECHNOLOGIES Duesseldorf GmbH, HRB 56063
Managing Director: Bo PENG, Qiuen PENG, Shengli WANG

Re: [PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Mimi Zohar]

Hi Dmitry,

On Fri, 2017-12-01 at 20:40 +0200, Dmitry Kasatkin wrote:
> The original design was discussed 3+ years ago, but was never 
> completed/upstreamed.
> Based on the recent discussions with Linus
> https://patchwork.kernel.org/patch/9975919, I've rebased this patch.
> 
> Before IMA appraisal was introduced, IMA was using own integrity cache
> lock along with i_mutex. process_measurement and ima_file_free took
> the iint->mutex first and then the i_mutex, while setxattr, chmod and
> chown took the locks in reverse order. To resolve the potential deadlock,
> i_mutex was moved to protect entire IMA functionality and the redundant
> iint->mutex was eliminated.
> 
> Solution was based on the assumption that filesystem code does not take
> i_mutex further. But when file is opened with O_DIRECT flag, direct-io
> implementation takes i_mutex and produces deadlock. Furthermore, certain
> other filesystem operations, such as llseek, also take i_mutex.
> 
> More recently some filesystems have replaced their filesystem specific
> lock with the global i_rwsem to read a file.  As a result, when IMA
> attempts to calculate the file hash, reading the file attempts to take
> the i_rwsem again.
> 
> To resolve O_DIRECT related deadlock problem, this patch re-introduces
> iint->mutex. But to eliminate the original chmod() related deadlock
> problem, this patch eliminates the requirement for chmod hooks to take
> the iint->mutex by introducing additional atomic iint->attr_flags to
> indicate calling of the hooks. The allowed locking order is to take
> the iint->mutex first and then the i_rwsem.
> 
> Original flags were cleared in chmod(), setxattr() or removwxattr() hooks
> and tested when file was closed or opened again. New atomic flags are set
> or cleared in those hooks and tested to clear iint->flags on close or on open.
> 
> Atomic flags are following:
> * IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, chown, chgrp)
>   and file attributes have changed. On file open, it causes IMA to clear
>   iint->flags to re-evaluate policy and perform IMA functions again.
> * IMA_CHANGE_XATTR - indicates that setxattr or removexattr was called and
>   extended attributes have changed. On file open, it causes IMA to clear
>   iint->flags IMA_DONE_MASK to re-appraise.
> * IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
>   It is cleared if file policy changes and no update is needed.
> * IMA_DIGSIG - indicates that file security.ima has signature and file
>   security.ima must not update to file has on file close.

Nice!  I've been testing with this patch and all seems good.  Before
queueing this patch to be upstreamed, I'd appreciate if others tested
using it as well.  It applies cleanly to the next-queued branch.

A subsequent patch will remove the O_DIRECT check in
ima_calc_file_hash().

Fixes: Commit 6552321831dc "xfs: remove i_iolock and use i_rwsem in
the VFS inode instead"

thanks,

Mimi

Re: [PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Mimi Zohar]

Hi Dmitry,

On Fri, 2017-12-01 at 20:40 +0200, Dmitry Kasatkin wrote:
> The original design was discussed 3+ years ago, but was never 
> completed/upstreamed.
> Based on the recent discussions with Linus
> https://patchwork.kernel.org/patch/9975919, I've rebased this patch.
> 
> Before IMA appraisal was introduced, IMA was using own integrity cache
> lock along with i_mutex. process_measurement and ima_file_free took
> the iint->mutex first and then the i_mutex, while setxattr, chmod and
> chown took the locks in reverse order. To resolve the potential deadlock,
> i_mutex was moved to protect entire IMA functionality and the redundant
> iint->mutex was eliminated.
> 
> Solution was based on the assumption that filesystem code does not take
> i_mutex further. But when file is opened with O_DIRECT flag, direct-io
> implementation takes i_mutex and produces deadlock. Furthermore, certain
> other filesystem operations, such as llseek, also take i_mutex.
> 
> More recently some filesystems have replaced their filesystem specific
> lock with the global i_rwsem to read a file.  As a result, when IMA
> attempts to calculate the file hash, reading the file attempts to take
> the i_rwsem again.
> 
> To resolve O_DIRECT related deadlock problem, this patch re-introduces
> iint->mutex. But to eliminate the original chmod() related deadlock
> problem, this patch eliminates the requirement for chmod hooks to take
> the iint->mutex by introducing additional atomic iint->attr_flags to
> indicate calling of the hooks. The allowed locking order is to take
> the iint->mutex first and then the i_rwsem.
> 
> Original flags were cleared in chmod(), setxattr() or removwxattr() hooks
> and tested when file was closed or opened again. New atomic flags are set
> or cleared in those hooks and tested to clear iint->flags on close or on open.
> 
> Atomic flags are following:
> * IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, chown, chgrp)
>   and file attributes have changed. On file open, it causes IMA to clear
>   iint->flags to re-evaluate policy and perform IMA functions again.
> * IMA_CHANGE_XATTR - indicates that setxattr or removexattr was called and
>   extended attributes have changed. On file open, it causes IMA to clear
>   iint->flags IMA_DONE_MASK to re-appraise.
> * IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
>   It is cleared if file policy changes and no update is needed.
> * IMA_DIGSIG - indicates that file security.ima has signature and file
>   security.ima must not update to file has on file close.

Nice!  I've been testing with this patch and all seems good.  Before
queueing this patch to be upstreamed, I'd appreciate if others tested
using it as well.  It applies cleanly to the next-queued branch.

A subsequent patch will remove the O_DIRECT check in
ima_calc_file_hash().

Fixes: Commit 6552321831dc "xfs: remove i_iolock and use i_rwsem in
the VFS inode instead"

thanks,

Mimi

[PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Dmitry Kasatkin]

The original design was discussed 3+ years ago, but was never 
completed/upstreamed.
Based on the recent discussions with Linus
https://patchwork.kernel.org/patch/9975919, I've rebased this patch.

Before IMA appraisal was introduced, IMA was using own integrity cache
lock along with i_mutex. process_measurement and ima_file_free took
the iint->mutex first and then the i_mutex, while setxattr, chmod and
chown took the locks in reverse order. To resolve the potential deadlock,
i_mutex was moved to protect entire IMA functionality and the redundant
iint->mutex was eliminated.

Solution was based on the assumption that filesystem code does not take
i_mutex further. But when file is opened with O_DIRECT flag, direct-io
implementation takes i_mutex and produces deadlock. Furthermore, certain
other filesystem operations, such as llseek, also take i_mutex.

More recently some filesystems have replaced their filesystem specific
lock with the global i_rwsem to read a file.  As a result, when IMA
attempts to calculate the file hash, reading the file attempts to take
the i_rwsem again.

To resolve O_DIRECT related deadlock problem, this patch re-introduces
iint->mutex. But to eliminate the original chmod() related deadlock
problem, this patch eliminates the requirement for chmod hooks to take
the iint->mutex by introducing additional atomic iint->attr_flags to
indicate calling of the hooks. The allowed locking order is to take
the iint->mutex first and then the i_rwsem.

Original flags were cleared in chmod(), setxattr() or removwxattr() hooks
and tested when file was closed or opened again. New atomic flags are set
or cleared in those hooks and tested to clear iint->flags on close or on open.

Atomic flags are following:
* IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, chown, chgrp)
  and file attributes have changed. On file open, it causes IMA to clear
  iint->flags to re-evaluate policy and perform IMA functions again.
* IMA_CHANGE_XATTR - indicates that setxattr or removexattr was called and
  extended attributes have changed. On file open, it causes IMA to clear
  iint->flags IMA_DONE_MASK to re-appraise.
* IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
  It is cleared if file policy changes and no update is needed.
* IMA_DIGSIG - indicates that file security.ima has signature and file
  security.ima must not update to file has on file close.

Changes in v5:
* use of inode_lock() and inode_unlock()

Changes in v4:
* adoped to violation detection fixes
* added IMA_UPDATE_XATTR flag to require xattr update on file close

Changes in v3:
* prevent signature removal with new locking
* rename attr_flags to atomic_flags

Changes in v2:
* revert taking the i_mutex in integrity_inode_get() so that iint allocation
  could be done with i_mutex taken
* move taking the i_mutex from appraisal code to the process_measurement()

Signed-off-by: Dmitry Kasatkin 
---
 security/integrity/iint.c |  2 ++
 security/integrity/ima/ima_appraise.c | 27 ---
 security/integrity/ima/ima_main.c | 65 ---
 security/integrity/integrity.h| 17 ++---
 4 files changed, 72 insertions(+), 39 deletions(-)

diff --git a/security/integrity/iint.c b/security/integrity/iint.c
index c84e058..d726ba23 100644
--- a/security/integrity/iint.c
+++ b/security/integrity/iint.c
@@ -155,12 +155,14 @@ static void init_once(void *foo)
memset(iint, 0, sizeof(*iint));
iint->version = 0;
iint->flags = 0UL;
+   iint->atomic_flags = 0;
iint->ima_file_status = INTEGRITY_UNKNOWN;
iint->ima_mmap_status = INTEGRITY_UNKNOWN;
iint->ima_bprm_status = INTEGRITY_UNKNOWN;
iint->ima_read_status = INTEGRITY_UNKNOWN;
iint->evm_status = INTEGRITY_UNKNOWN;
iint->measured_pcrs = 0;
+   mutex_init(>mutex);
 }
 
 static int __init integrity_iintcache_init(void)
diff --git a/security/integrity/ima/ima_appraise.c 
b/security/integrity/ima/ima_appraise.c
index 9a54c77..3fc96dbd 100644
--- a/security/integrity/ima/ima_appraise.c
+++ b/security/integrity/ima/ima_appraise.c
@@ -251,6 +251,7 @@ int ima_appraise_measurement(enum ima_hooks func,
status = INTEGRITY_FAIL;
break;
}
+   clear_bit(IMA_DIGSIG, >atomic_flags);
if (xattr_len - sizeof(xattr_value->type) - hash_start >=
iint->ima_hash->length)
/* xattr length may be longer. md5 hash in previous
@@ -269,7 +270,7 @@ int ima_appraise_measurement(enum ima_hooks func,
status = INTEGRITY_PASS;
break;
case EVM_IMA_XATTR_DIGSIG:
-   iint->flags |= IMA_DIGSIG;
+   set_bit(IMA_DIGSIG, >atomic_flags);
rc = integrity_digsig_verify(INTEGRITY_KEYRING_IMA,
 (const char 

[PATCHv5 1/1] ima: re-introduce own integrity cache lock

[Dmitry Kasatkin]

The original design was discussed 3+ years ago, but was never 
completed/upstreamed.
Based on the recent discussions with Linus
https://patchwork.kernel.org/patch/9975919, I've rebased this patch.

Before IMA appraisal was introduced, IMA was using own integrity cache
lock along with i_mutex. process_measurement and ima_file_free took
the iint->mutex first and then the i_mutex, while setxattr, chmod and
chown took the locks in reverse order. To resolve the potential deadlock,
i_mutex was moved to protect entire IMA functionality and the redundant
iint->mutex was eliminated.

Solution was based on the assumption that filesystem code does not take
i_mutex further. But when file is opened with O_DIRECT flag, direct-io
implementation takes i_mutex and produces deadlock. Furthermore, certain
other filesystem operations, such as llseek, also take i_mutex.

More recently some filesystems have replaced their filesystem specific
lock with the global i_rwsem to read a file.  As a result, when IMA
attempts to calculate the file hash, reading the file attempts to take
the i_rwsem again.

To resolve O_DIRECT related deadlock problem, this patch re-introduces
iint->mutex. But to eliminate the original chmod() related deadlock
problem, this patch eliminates the requirement for chmod hooks to take
the iint->mutex by introducing additional atomic iint->attr_flags to
indicate calling of the hooks. The allowed locking order is to take
the iint->mutex first and then the i_rwsem.

Original flags were cleared in chmod(), setxattr() or removwxattr() hooks
and tested when file was closed or opened again. New atomic flags are set
or cleared in those hooks and tested to clear iint->flags on close or on open.

Atomic flags are following:
* IMA_CHANGE_ATTR - indicates that chATTR() was called (chmod, chown, chgrp)
  and file attributes have changed. On file open, it causes IMA to clear
  iint->flags to re-evaluate policy and perform IMA functions again.
* IMA_CHANGE_XATTR - indicates that setxattr or removexattr was called and
  extended attributes have changed. On file open, it causes IMA to clear
  iint->flags IMA_DONE_MASK to re-appraise.
* IMA_UPDATE_XATTR - indicates that security.ima needs to be updated.
  It is cleared if file policy changes and no update is needed.
* IMA_DIGSIG - indicates that file security.ima has signature and file
  security.ima must not update to file has on file close.

Changes in v5:
* use of inode_lock() and inode_unlock()

Changes in v4:
* adoped to violation detection fixes
* added IMA_UPDATE_XATTR flag to require xattr update on file close

Changes in v3:
* prevent signature removal with new locking
* rename attr_flags to atomic_flags

Changes in v2:
* revert taking the i_mutex in integrity_inode_get() so that iint allocation
  could be done with i_mutex taken
* move taking the i_mutex from appraisal code to the process_measurement()

Signed-off-by: Dmitry Kasatkin 
---
 security/integrity/iint.c |  2 ++
 security/integrity/ima/ima_appraise.c | 27 ---
 security/integrity/ima/ima_main.c | 65 ---
 security/integrity/integrity.h| 17 ++---
 4 files changed, 72 insertions(+), 39 deletions(-)

diff --git a/security/integrity/iint.c b/security/integrity/iint.c
index c84e058..d726ba23 100644
--- a/security/integrity/iint.c
+++ b/security/integrity/iint.c
@@ -155,12 +155,14 @@ static void init_once(void *foo)
memset(iint, 0, sizeof(*iint));
iint->version = 0;
iint->flags = 0UL;
+   iint->atomic_flags = 0;
iint->ima_file_status = INTEGRITY_UNKNOWN;
iint->ima_mmap_status = INTEGRITY_UNKNOWN;
iint->ima_bprm_status = INTEGRITY_UNKNOWN;
iint->ima_read_status = INTEGRITY_UNKNOWN;
iint->evm_status = INTEGRITY_UNKNOWN;
iint->measured_pcrs = 0;
+   mutex_init(>mutex);
 }
 
 static int __init integrity_iintcache_init(void)
diff --git a/security/integrity/ima/ima_appraise.c 
b/security/integrity/ima/ima_appraise.c
index 9a54c77..3fc96dbd 100644
--- a/security/integrity/ima/ima_appraise.c
+++ b/security/integrity/ima/ima_appraise.c
@@ -251,6 +251,7 @@ int ima_appraise_measurement(enum ima_hooks func,
status = INTEGRITY_FAIL;
break;
}
+   clear_bit(IMA_DIGSIG, >atomic_flags);
if (xattr_len - sizeof(xattr_value->type) - hash_start >=
iint->ima_hash->length)
/* xattr length may be longer. md5 hash in previous
@@ -269,7 +270,7 @@ int ima_appraise_measurement(enum ima_hooks func,
status = INTEGRITY_PASS;
break;
case EVM_IMA_XATTR_DIGSIG:
-   iint->flags |= IMA_DIGSIG;
+   set_bit(IMA_DIGSIG, >atomic_flags);
rc = integrity_digsig_verify(INTEGRITY_KEYRING_IMA,
 (const char *)xattr_value, rc,