2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-06-28 03:21:41 -07:00
|
|
|
|
|
|
|
|
|
2019-10-08 14:41:55 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-02-08 18:51:29 +01:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
2017-03-06 16:21:16 +02:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-03-09 18:07:34 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
2015-06-17 15:28:02 +01:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2018-07-20 12:01:48 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-07 12:15:45 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2019-11-07 12:15:45 +01:00
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2019-11-07 12:15:45 +01:00
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2011-02-03 12:18:59 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2012-10-24 12:39:02 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-10-24 12:39:02 +01:00
|
|
|
|
2015-01-08 18:06:01 +00:00
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
|
|
|
|
|
2019-11-07 12:15:46 +01:00
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
2019-11-07 12:15:46 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2012-10-24 12:39:02 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-10 14:56:42 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2026-02-21 16:37:42 -08:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-02-21 16:37:42 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2026-02-21 16:37:42 -08:00
|
|
|
|
2010-12-10 14:56:42 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
2026-02-21 20:03:00 -08:00
|
|
|
|
|
|
|
|
|
2021-03-10 11:45:00 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-02-21 16:37:42 -08:00
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-01-08 18:06:01 +00:00
|
|
|
|
2011-03-09 18:07:34 -05:00
|
|
|
|
|
|
|
|
|
2010-12-10 14:56:42 +00:00
|
|
|
|
2021-03-10 11:46:13 +01:00
|
|
|
|
|
|
|
|
|
2021-03-10 11:45:26 +01:00
|
|
|
|
|
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
2021-03-10 11:45:26 +01:00
|
|
|
|
|
|
|
|
|
2021-03-10 11:45:00 +01:00
|
|
|
|
2010-12-10 14:56:42 +00:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2017-03-06 16:21:16 +02:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-10-24 12:39:02 +01:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-10 14:56:42 +00:00
|
|
|
|
2017-03-06 16:21:16 +02:00
|
|
|
|
2011-02-03 12:19:01 -05:00
|
|
|
|
|
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-12-10 11:28:17 +02:00
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
|
|
|
|
|
2011-10-27 17:58:49 -04:00
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
2011-10-27 17:58:49 -04:00
|
|
|
|
|
|
|
|
|
2012-10-24 12:39:02 +01:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-07-11 20:58:43 -07:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
2021-07-30 09:18:04 +02:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-08 10:54:32 -08:00
|
|
|
|
|
|
|
|
|
2025-08-11 13:26:31 +02:00
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
2021-03-10 11:45:26 +01:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
2011-02-09 16:11:32 -05:00
|
|
|
|
2021-03-10 11:45:26 +01:00
|
|
|
|
2011-02-09 16:11:32 -05:00
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
2011-03-08 16:56:43 +00:00
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-03-10 11:45:26 +01:00
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
2011-09-29 11:57:56 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-15 08:51:07 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-09-29 11:57:56 +01:00
|
|
|
|
2021-02-15 08:51:07 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-09-29 11:57:56 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-15 08:51:07 +01:00
|
|
|
|
2011-09-29 11:57:56 +01:00
|
|
|
|
|
|
|
|
|
2015-01-05 14:13:41 +00:00
|
|
|
|
2021-03-10 11:45:26 +01:00
|
|
|
|
2011-09-29 11:57:56 +01:00
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2021-03-10 11:45:00 +01:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2021-02-15 08:52:27 +01:00
|
|
|
|
xen/gntdev: Prevent leaking grants
Prior to this commit, if a grant mapping operation failed partially,
some of the entries in the map_ops array would be invalid, whereas all
of the entries in the kmap_ops array would be valid. This in turn would
cause the following logic in gntdev_map_grant_pages to become invalid:
for (i = 0; i < map->count; i++) {
if (map->map_ops[i].status == GNTST_okay) {
map->unmap_ops[i].handle = map->map_ops[i].handle;
if (!use_ptemod)
alloced++;
}
if (use_ptemod) {
if (map->kmap_ops[i].status == GNTST_okay) {
if (map->map_ops[i].status == GNTST_okay)
alloced++;
map->kunmap_ops[i].handle = map->kmap_ops[i].handle;
}
}
}
...
atomic_add(alloced, &map->live_grants);
Assume that use_ptemod is true (i.e., the domain mapping the granted
pages is a paravirtualized domain). In the code excerpt above, note that
the "alloced" variable is only incremented when both kmap_ops[i].status
and map_ops[i].status are set to GNTST_okay (i.e., both mapping
operations are successful). However, as also noted above, there are
cases where a grant mapping operation fails partially, breaking the
assumption of the code excerpt above.
The aforementioned causes map->live_grants to be incorrectly set. In
some cases, all of the map_ops mappings fail, but all of the kmap_ops
mappings succeed, meaning that live_grants may remain zero. This in turn
makes it impossible to unmap the successfully grant-mapped pages pointed
to by kmap_ops, because unmap_grant_pages has the following snippet of
code at its beginning:
if (atomic_read(&map->live_grants) == 0)
return; /* Nothing to do */
In other cases where only some of the map_ops mappings fail but all
kmap_ops mappings succeed, live_grants is made positive, but when the
user requests unmapping the grant-mapped pages, __unmap_grant_pages_done
will then make map->live_grants negative, because the latter function
does not check if all of the pages that were requested to be unmapped
were actually unmapped, and the same function unconditionally subtracts
"data->count" (i.e., a value that can be greater than map->live_grants)
from map->live_grants. The side effects of a negative live_grants value
have not been studied.
The net effect of all of this is that grant references are leaked in one
of the above conditions. In Qubes OS v4.1 (which uses Xen's grant
mechanism extensively for X11 GUI isolation), this issue manifests
itself with warning messages like the following to be printed out by the
Linux kernel in the VM that had granted pages (that contain X11 GUI
window data) to dom0: "g.e. 0x1234 still pending", especially after the
user rapidly resizes GUI VM windows (causing some grant-mapping
operations to partially or completely fail, due to the fact that the VM
unshares some of the pages as part of the window resizing, making the
pages impossible to grant-map from dom0).
The fix for this issue involves counting all successful map_ops and
kmap_ops mappings separately, and then adding the sum to live_grants.
During unmapping, only the number of successfully unmapped grants is
subtracted from live_grants. The code is also modified to check for
negative live_grants values after the subtraction and warn the user.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: dbe97cff7dd9 ("xen/gntdev: Avoid blocking in unmap_grant_pages()")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Acked-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-2-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:05 -04:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2015-01-05 14:13:41 +00:00
|
|
|
|
2021-02-15 08:51:07 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
xen/gntdev: Prevent leaking grants
Prior to this commit, if a grant mapping operation failed partially,
some of the entries in the map_ops array would be invalid, whereas all
of the entries in the kmap_ops array would be valid. This in turn would
cause the following logic in gntdev_map_grant_pages to become invalid:
for (i = 0; i < map->count; i++) {
if (map->map_ops[i].status == GNTST_okay) {
map->unmap_ops[i].handle = map->map_ops[i].handle;
if (!use_ptemod)
alloced++;
}
if (use_ptemod) {
if (map->kmap_ops[i].status == GNTST_okay) {
if (map->map_ops[i].status == GNTST_okay)
alloced++;
map->kunmap_ops[i].handle = map->kmap_ops[i].handle;
}
}
}
...
atomic_add(alloced, &map->live_grants);
Assume that use_ptemod is true (i.e., the domain mapping the granted
pages is a paravirtualized domain). In the code excerpt above, note that
the "alloced" variable is only incremented when both kmap_ops[i].status
and map_ops[i].status are set to GNTST_okay (i.e., both mapping
operations are successful). However, as also noted above, there are
cases where a grant mapping operation fails partially, breaking the
assumption of the code excerpt above.
The aforementioned causes map->live_grants to be incorrectly set. In
some cases, all of the map_ops mappings fail, but all of the kmap_ops
mappings succeed, meaning that live_grants may remain zero. This in turn
makes it impossible to unmap the successfully grant-mapped pages pointed
to by kmap_ops, because unmap_grant_pages has the following snippet of
code at its beginning:
if (atomic_read(&map->live_grants) == 0)
return; /* Nothing to do */
In other cases where only some of the map_ops mappings fail but all
kmap_ops mappings succeed, live_grants is made positive, but when the
user requests unmapping the grant-mapped pages, __unmap_grant_pages_done
will then make map->live_grants negative, because the latter function
does not check if all of the pages that were requested to be unmapped
were actually unmapped, and the same function unconditionally subtracts
"data->count" (i.e., a value that can be greater than map->live_grants)
from map->live_grants. The side effects of a negative live_grants value
have not been studied.
The net effect of all of this is that grant references are leaked in one
of the above conditions. In Qubes OS v4.1 (which uses Xen's grant
mechanism extensively for X11 GUI isolation), this issue manifests
itself with warning messages like the following to be printed out by the
Linux kernel in the VM that had granted pages (that contain X11 GUI
window data) to dom0: "g.e. 0x1234 still pending", especially after the
user rapidly resizes GUI VM windows (causing some grant-mapping
operations to partially or completely fail, due to the fact that the VM
unshares some of the pages as part of the window resizing, making the
pages impossible to grant-map from dom0).
The fix for this issue involves counting all successful map_ops and
kmap_ops mappings separately, and then adding the sum to live_grants.
During unmapping, only the number of successfully unmapped grants is
subtracted from live_grants. The code is also modified to check for
negative live_grants values after the subtraction and warn the user.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: dbe97cff7dd9 ("xen/gntdev: Avoid blocking in unmap_grant_pages()")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Acked-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-2-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:05 -04:00
|
|
|
|
2021-02-15 08:52:27 +01:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2021-02-15 08:52:27 +01:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Prevent leaking grants
Prior to this commit, if a grant mapping operation failed partially,
some of the entries in the map_ops array would be invalid, whereas all
of the entries in the kmap_ops array would be valid. This in turn would
cause the following logic in gntdev_map_grant_pages to become invalid:
for (i = 0; i < map->count; i++) {
if (map->map_ops[i].status == GNTST_okay) {
map->unmap_ops[i].handle = map->map_ops[i].handle;
if (!use_ptemod)
alloced++;
}
if (use_ptemod) {
if (map->kmap_ops[i].status == GNTST_okay) {
if (map->map_ops[i].status == GNTST_okay)
alloced++;
map->kunmap_ops[i].handle = map->kmap_ops[i].handle;
}
}
}
...
atomic_add(alloced, &map->live_grants);
Assume that use_ptemod is true (i.e., the domain mapping the granted
pages is a paravirtualized domain). In the code excerpt above, note that
the "alloced" variable is only incremented when both kmap_ops[i].status
and map_ops[i].status are set to GNTST_okay (i.e., both mapping
operations are successful). However, as also noted above, there are
cases where a grant mapping operation fails partially, breaking the
assumption of the code excerpt above.
The aforementioned causes map->live_grants to be incorrectly set. In
some cases, all of the map_ops mappings fail, but all of the kmap_ops
mappings succeed, meaning that live_grants may remain zero. This in turn
makes it impossible to unmap the successfully grant-mapped pages pointed
to by kmap_ops, because unmap_grant_pages has the following snippet of
code at its beginning:
if (atomic_read(&map->live_grants) == 0)
return; /* Nothing to do */
In other cases where only some of the map_ops mappings fail but all
kmap_ops mappings succeed, live_grants is made positive, but when the
user requests unmapping the grant-mapped pages, __unmap_grant_pages_done
will then make map->live_grants negative, because the latter function
does not check if all of the pages that were requested to be unmapped
were actually unmapped, and the same function unconditionally subtracts
"data->count" (i.e., a value that can be greater than map->live_grants)
from map->live_grants. The side effects of a negative live_grants value
have not been studied.
The net effect of all of this is that grant references are leaked in one
of the above conditions. In Qubes OS v4.1 (which uses Xen's grant
mechanism extensively for X11 GUI isolation), this issue manifests
itself with warning messages like the following to be printed out by the
Linux kernel in the VM that had granted pages (that contain X11 GUI
window data) to dom0: "g.e. 0x1234 still pending", especially after the
user rapidly resizes GUI VM windows (causing some grant-mapping
operations to partially or completely fail, due to the fact that the VM
unshares some of the pages as part of the window resizing, making the
pages impossible to grant-map from dom0).
The fix for this issue involves counting all successful map_ops and
kmap_ops mappings separately, and then adding the sum to live_grants.
During unmapping, only the number of successfully unmapped grants is
subtracted from live_grants. The code is also modified to check for
negative live_grants values after the subtraction and warn the user.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: dbe97cff7dd9 ("xen/gntdev: Avoid blocking in unmap_grant_pages()")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Acked-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-2-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:05 -04:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
xen/gntdev: Prevent leaking grants
Prior to this commit, if a grant mapping operation failed partially,
some of the entries in the map_ops array would be invalid, whereas all
of the entries in the kmap_ops array would be valid. This in turn would
cause the following logic in gntdev_map_grant_pages to become invalid:
for (i = 0; i < map->count; i++) {
if (map->map_ops[i].status == GNTST_okay) {
map->unmap_ops[i].handle = map->map_ops[i].handle;
if (!use_ptemod)
alloced++;
}
if (use_ptemod) {
if (map->kmap_ops[i].status == GNTST_okay) {
if (map->map_ops[i].status == GNTST_okay)
alloced++;
map->kunmap_ops[i].handle = map->kmap_ops[i].handle;
}
}
}
...
atomic_add(alloced, &map->live_grants);
Assume that use_ptemod is true (i.e., the domain mapping the granted
pages is a paravirtualized domain). In the code excerpt above, note that
the "alloced" variable is only incremented when both kmap_ops[i].status
and map_ops[i].status are set to GNTST_okay (i.e., both mapping
operations are successful). However, as also noted above, there are
cases where a grant mapping operation fails partially, breaking the
assumption of the code excerpt above.
The aforementioned causes map->live_grants to be incorrectly set. In
some cases, all of the map_ops mappings fail, but all of the kmap_ops
mappings succeed, meaning that live_grants may remain zero. This in turn
makes it impossible to unmap the successfully grant-mapped pages pointed
to by kmap_ops, because unmap_grant_pages has the following snippet of
code at its beginning:
if (atomic_read(&map->live_grants) == 0)
return; /* Nothing to do */
In other cases where only some of the map_ops mappings fail but all
kmap_ops mappings succeed, live_grants is made positive, but when the
user requests unmapping the grant-mapped pages, __unmap_grant_pages_done
will then make map->live_grants negative, because the latter function
does not check if all of the pages that were requested to be unmapped
were actually unmapped, and the same function unconditionally subtracts
"data->count" (i.e., a value that can be greater than map->live_grants)
from map->live_grants. The side effects of a negative live_grants value
have not been studied.
The net effect of all of this is that grant references are leaked in one
of the above conditions. In Qubes OS v4.1 (which uses Xen's grant
mechanism extensively for X11 GUI isolation), this issue manifests
itself with warning messages like the following to be printed out by the
Linux kernel in the VM that had granted pages (that contain X11 GUI
window data) to dom0: "g.e. 0x1234 still pending", especially after the
user rapidly resizes GUI VM windows (causing some grant-mapping
operations to partially or completely fail, due to the fact that the VM
unshares some of the pages as part of the window resizing, making the
pages impossible to grant-map from dom0).
The fix for this issue involves counting all successful map_ops and
kmap_ops mappings separately, and then adding the sum to live_grants.
During unmapping, only the number of successfully unmapped grants is
subtracted from live_grants. The code is also modified to check for
negative live_grants values after the subtraction and warn the user.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: dbe97cff7dd9 ("xen/gntdev: Avoid blocking in unmap_grant_pages()")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Acked-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-2-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:05 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-07-10 19:05:22 -04:00
|
|
|
|
|
|
|
|
|
2011-02-23 08:11:35 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-03-10 11:45:26 +01:00
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
xen/gntdev: Prevent leaking grants
Prior to this commit, if a grant mapping operation failed partially,
some of the entries in the map_ops array would be invalid, whereas all
of the entries in the kmap_ops array would be valid. This in turn would
cause the following logic in gntdev_map_grant_pages to become invalid:
for (i = 0; i < map->count; i++) {
if (map->map_ops[i].status == GNTST_okay) {
map->unmap_ops[i].handle = map->map_ops[i].handle;
if (!use_ptemod)
alloced++;
}
if (use_ptemod) {
if (map->kmap_ops[i].status == GNTST_okay) {
if (map->map_ops[i].status == GNTST_okay)
alloced++;
map->kunmap_ops[i].handle = map->kmap_ops[i].handle;
}
}
}
...
atomic_add(alloced, &map->live_grants);
Assume that use_ptemod is true (i.e., the domain mapping the granted
pages is a paravirtualized domain). In the code excerpt above, note that
the "alloced" variable is only incremented when both kmap_ops[i].status
and map_ops[i].status are set to GNTST_okay (i.e., both mapping
operations are successful). However, as also noted above, there are
cases where a grant mapping operation fails partially, breaking the
assumption of the code excerpt above.
The aforementioned causes map->live_grants to be incorrectly set. In
some cases, all of the map_ops mappings fail, but all of the kmap_ops
mappings succeed, meaning that live_grants may remain zero. This in turn
makes it impossible to unmap the successfully grant-mapped pages pointed
to by kmap_ops, because unmap_grant_pages has the following snippet of
code at its beginning:
if (atomic_read(&map->live_grants) == 0)
return; /* Nothing to do */
In other cases where only some of the map_ops mappings fail but all
kmap_ops mappings succeed, live_grants is made positive, but when the
user requests unmapping the grant-mapped pages, __unmap_grant_pages_done
will then make map->live_grants negative, because the latter function
does not check if all of the pages that were requested to be unmapped
were actually unmapped, and the same function unconditionally subtracts
"data->count" (i.e., a value that can be greater than map->live_grants)
from map->live_grants. The side effects of a negative live_grants value
have not been studied.
The net effect of all of this is that grant references are leaked in one
of the above conditions. In Qubes OS v4.1 (which uses Xen's grant
mechanism extensively for X11 GUI isolation), this issue manifests
itself with warning messages like the following to be printed out by the
Linux kernel in the VM that had granted pages (that contain X11 GUI
window data) to dom0: "g.e. 0x1234 still pending", especially after the
user rapidly resizes GUI VM windows (causing some grant-mapping
operations to partially or completely fail, due to the fact that the VM
unshares some of the pages as part of the window resizing, making the
pages impossible to grant-map from dom0).
The fix for this issue involves counting all successful map_ops and
kmap_ops mappings separately, and then adding the sum to live_grants.
During unmapping, only the number of successfully unmapped grants is
subtracted from live_grants. The code is also modified to check for
negative live_grants values after the subtraction and warn the user.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: dbe97cff7dd9 ("xen/gntdev: Avoid blocking in unmap_grant_pages()")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Acked-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-2-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:05 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-07-10 19:05:22 -04:00
|
|
|
|
|
|
|
|
|
2021-09-17 08:13:08 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
xen/gntdev: Prevent leaking grants
Prior to this commit, if a grant mapping operation failed partially,
some of the entries in the map_ops array would be invalid, whereas all
of the entries in the kmap_ops array would be valid. This in turn would
cause the following logic in gntdev_map_grant_pages to become invalid:
for (i = 0; i < map->count; i++) {
if (map->map_ops[i].status == GNTST_okay) {
map->unmap_ops[i].handle = map->map_ops[i].handle;
if (!use_ptemod)
alloced++;
}
if (use_ptemod) {
if (map->kmap_ops[i].status == GNTST_okay) {
if (map->map_ops[i].status == GNTST_okay)
alloced++;
map->kunmap_ops[i].handle = map->kmap_ops[i].handle;
}
}
}
...
atomic_add(alloced, &map->live_grants);
Assume that use_ptemod is true (i.e., the domain mapping the granted
pages is a paravirtualized domain). In the code excerpt above, note that
the "alloced" variable is only incremented when both kmap_ops[i].status
and map_ops[i].status are set to GNTST_okay (i.e., both mapping
operations are successful). However, as also noted above, there are
cases where a grant mapping operation fails partially, breaking the
assumption of the code excerpt above.
The aforementioned causes map->live_grants to be incorrectly set. In
some cases, all of the map_ops mappings fail, but all of the kmap_ops
mappings succeed, meaning that live_grants may remain zero. This in turn
makes it impossible to unmap the successfully grant-mapped pages pointed
to by kmap_ops, because unmap_grant_pages has the following snippet of
code at its beginning:
if (atomic_read(&map->live_grants) == 0)
return; /* Nothing to do */
In other cases where only some of the map_ops mappings fail but all
kmap_ops mappings succeed, live_grants is made positive, but when the
user requests unmapping the grant-mapped pages, __unmap_grant_pages_done
will then make map->live_grants negative, because the latter function
does not check if all of the pages that were requested to be unmapped
were actually unmapped, and the same function unconditionally subtracts
"data->count" (i.e., a value that can be greater than map->live_grants)
from map->live_grants. The side effects of a negative live_grants value
have not been studied.
The net effect of all of this is that grant references are leaked in one
of the above conditions. In Qubes OS v4.1 (which uses Xen's grant
mechanism extensively for X11 GUI isolation), this issue manifests
itself with warning messages like the following to be printed out by the
Linux kernel in the VM that had granted pages (that contain X11 GUI
window data) to dom0: "g.e. 0x1234 still pending", especially after the
user rapidly resizes GUI VM windows (causing some grant-mapping
operations to partially or completely fail, due to the fact that the VM
unshares some of the pages as part of the window resizing, making the
pages impossible to grant-map from dom0).
The fix for this issue involves counting all successful map_ops and
kmap_ops mappings separately, and then adding the sum to live_grants.
During unmapping, only the number of successfully unmapped grants is
subtracted from live_grants. The code is also modified to check for
negative live_grants values after the subtraction and warn the user.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: dbe97cff7dd9 ("xen/gntdev: Avoid blocking in unmap_grant_pages()")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Acked-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-2-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:05 -04:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Prevent leaking grants
Prior to this commit, if a grant mapping operation failed partially,
some of the entries in the map_ops array would be invalid, whereas all
of the entries in the kmap_ops array would be valid. This in turn would
cause the following logic in gntdev_map_grant_pages to become invalid:
for (i = 0; i < map->count; i++) {
if (map->map_ops[i].status == GNTST_okay) {
map->unmap_ops[i].handle = map->map_ops[i].handle;
if (!use_ptemod)
alloced++;
}
if (use_ptemod) {
if (map->kmap_ops[i].status == GNTST_okay) {
if (map->map_ops[i].status == GNTST_okay)
alloced++;
map->kunmap_ops[i].handle = map->kmap_ops[i].handle;
}
}
}
...
atomic_add(alloced, &map->live_grants);
Assume that use_ptemod is true (i.e., the domain mapping the granted
pages is a paravirtualized domain). In the code excerpt above, note that
the "alloced" variable is only incremented when both kmap_ops[i].status
and map_ops[i].status are set to GNTST_okay (i.e., both mapping
operations are successful). However, as also noted above, there are
cases where a grant mapping operation fails partially, breaking the
assumption of the code excerpt above.
The aforementioned causes map->live_grants to be incorrectly set. In
some cases, all of the map_ops mappings fail, but all of the kmap_ops
mappings succeed, meaning that live_grants may remain zero. This in turn
makes it impossible to unmap the successfully grant-mapped pages pointed
to by kmap_ops, because unmap_grant_pages has the following snippet of
code at its beginning:
if (atomic_read(&map->live_grants) == 0)
return; /* Nothing to do */
In other cases where only some of the map_ops mappings fail but all
kmap_ops mappings succeed, live_grants is made positive, but when the
user requests unmapping the grant-mapped pages, __unmap_grant_pages_done
will then make map->live_grants negative, because the latter function
does not check if all of the pages that were requested to be unmapped
were actually unmapped, and the same function unconditionally subtracts
"data->count" (i.e., a value that can be greater than map->live_grants)
from map->live_grants. The side effects of a negative live_grants value
have not been studied.
The net effect of all of this is that grant references are leaked in one
of the above conditions. In Qubes OS v4.1 (which uses Xen's grant
mechanism extensively for X11 GUI isolation), this issue manifests
itself with warning messages like the following to be printed out by the
Linux kernel in the VM that had granted pages (that contain X11 GUI
window data) to dom0: "g.e. 0x1234 still pending", especially after the
user rapidly resizes GUI VM windows (causing some grant-mapping
operations to partially or completely fail, due to the fact that the VM
unshares some of the pages as part of the window resizing, making the
pages impossible to grant-map from dom0).
The fix for this issue involves counting all successful map_ops and
kmap_ops mappings separately, and then adding the sum to live_grants.
During unmapping, only the number of successfully unmapped grants is
subtracted from live_grants. The code is also modified to check for
negative live_grants values after the subtraction and warn the user.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: dbe97cff7dd9 ("xen/gntdev: Avoid blocking in unmap_grant_pages()")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Acked-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-2-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:05 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
2011-02-09 15:12:00 -05:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-09 15:12:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
2011-02-09 15:12:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2011-02-09 15:12:00 -05:00
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
2011-02-09 15:12:00 -05:00
|
|
|
|
|
|
|
|
|
xen/gntdev: Avoid blocking in unmap_grant_pages()
unmap_grant_pages() currently waits for the pages to no longer be used.
In https://github.com/QubesOS/qubes-issues/issues/7481, this lead to a
deadlock against i915: i915 was waiting for gntdev's MMU notifier to
finish, while gntdev was waiting for i915 to free its pages. I also
believe this is responsible for various deadlocks I have experienced in
the past.
Avoid these problems by making unmap_grant_pages async. This requires
making it return void, as any errors will not be available when the
function returns. Fortunately, the only use of the return value is a
WARN_ON(), which can be replaced by a WARN_ON when the error is
detected. Additionally, a failed call will not prevent further calls
from being made, but this is harmless.
Because unmap_grant_pages is now async, the grant handle will be sent to
INVALID_GRANT_HANDLE too late to prevent multiple unmaps of the same
handle. Instead, a separate bool array is allocated for this purpose.
This wastes memory, but stuffing this information in padding bytes is
too fragile. Furthermore, it is necessary to grab a reference to the
map before making the asynchronous call, and release the reference when
the call returns.
It is also necessary to guard against reentrancy in gntdev_map_put(),
and to handle the case where userspace tries to map a mapping whose
contents have not all been freed yet.
Fixes: 745282256c75 ("xen/gntdev: safely unmap grants in case they are still in use")
Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20220622022726.2538-1-demi@invisiblethingslab.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-06-21 22:27:26 -04:00
|
|
|
|
|
|
|
|
|
2011-02-09 15:12:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2011-03-07 15:18:57 -05:00
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2011-03-07 15:18:57 -05:00
|
|
|
|
|
|
|
|
|
2017-03-06 16:21:16 +02:00
|
|
|
|
2011-03-07 15:18:57 -05:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2011-03-07 15:18:57 -05:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2025-08-11 13:26:31 +02:00
|
|
|
|
2014-12-18 14:59:07 +00:00
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2014-12-18 14:59:07 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-09-09 15:39:26 -07:00
|
|
|
|
2011-03-07 15:18:57 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2025-08-11 13:26:31 +02:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
|
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
|
|
|
|
|
2018-09-05 09:21:39 +10:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
2018-09-05 09:21:39 +10:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
|
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-09-05 09:21:39 +10:00
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
|
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-02-21 16:37:42 -08:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:48 +03:00
|
|
|
|
2019-02-14 16:23:20 +02:00
|
|
|
|
2018-07-20 12:01:48 +03:00
|
|
|
|
2019-11-11 12:20:09 +00:00
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:48 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
|
|
|
|
|
2019-10-08 14:41:55 -05:00
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-06-26 03:28:24 +02:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:01 -05:00
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2015-06-26 03:28:24 +02:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:48 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-07 12:15:45 +01:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:46 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2011-02-03 12:18:59 -05:00
|
|
|
|
2011-02-03 12:19:01 -05:00
|
|
|
|
|
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
|
|
|
|
|
2011-02-03 12:18:59 -05:00
|
|
|
|
|
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-06-22 19:25:41 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:01 -05:00
|
|
|
|
2026-06-22 19:25:41 +08:00
|
|
|
|
2011-02-03 12:19:01 -05:00
|
|
|
|
2026-06-22 19:25:41 +08:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2011-02-03 12:19:01 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2011-10-11 15:16:06 -04:00
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:00 -05:00
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2013-01-02 22:57:11 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-06-08 21:33:25 -07:00
|
|
|
|
2011-02-03 12:19:00 -05:00
|
|
|
|
|
|
|
|
|
2013-01-02 22:57:11 +00:00
|
|
|
|
2011-02-03 12:19:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-01-02 22:57:11 +00:00
|
|
|
|
2011-02-03 12:19:00 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2013-01-02 22:57:11 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2013-01-02 22:57:11 +00:00
|
|
|
|
2020-06-08 21:33:25 -07:00
|
|
|
|
2013-01-02 22:57:11 +00:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2013-01-02 22:57:11 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
2011-10-27 17:58:49 -04:00
|
|
|
|
2020-03-23 18:15:11 +02:00
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-10-27 17:58:49 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-09 18:15:50 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-10-27 17:58:49 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-10-27 17:58:49 -04:00
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
2011-10-27 17:58:49 -04:00
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2011-10-27 17:58:49 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
2020-09-06 12:21:53 +05:30
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-09-06 12:21:54 +05:30
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-09-06 12:21:54 +05:30
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
2020-09-06 12:21:53 +05:30
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-09-06 12:21:53 +05:30
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
2020-09-06 12:21:53 +05:30
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-09-06 12:21:53 +05:30
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
2020-09-06 12:21:53 +05:30
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
|
|
|
|
|
2026-02-21 16:37:42 -08:00
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-07-03 09:32:59 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:04 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-12-02 16:13:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:48 +03:00
|
|
|
|
|
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
2018-07-20 12:01:48 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-05-24 05:34:32 +05:30
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2019-05-13 17:22:23 -07:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
2017-03-06 16:21:16 +02:00
|
|
|
|
2011-02-03 12:19:01 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2023-01-26 11:37:49 -08:00
|
|
|
|
2011-03-07 15:18:57 -05:00
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
2023-01-26 11:37:49 -08:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2011-02-09 16:11:32 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-03-19 08:45:43 +03:00
|
|
|
|
2011-02-09 16:11:32 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
|
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
xen/gntdev: Accommodate VMA splitting
Prior to this commit, the gntdev driver code did not handle the
following scenario correctly with paravirtualized (PV) Xen domains:
* User process sets up a gntdev mapping composed of two grant mappings
(i.e., two pages shared by another Xen domain).
* User process munmap()s one of the pages.
* User process munmap()s the remaining page.
* User process exits.
In the scenario above, the user process would cause the kernel to log
the following messages in dmesg for the first munmap(), and the second
munmap() call would result in similar log messages:
BUG: Bad page map in process doublemap.test pte:... pmd:...
page:0000000057c97bff refcount:1 mapcount:-1 \
mapping:0000000000000000 index:0x0 pfn:...
...
page dumped because: bad pte
...
file:gntdev fault:0x0 mmap:gntdev_mmap [xen_gntdev] readpage:0x0
...
Call Trace:
<TASK>
dump_stack_lvl+0x46/0x5e
print_bad_pte.cold+0x66/0xb6
unmap_page_range+0x7e5/0xdc0
unmap_vmas+0x78/0xf0
unmap_region+0xa8/0x110
__do_munmap+0x1ea/0x4e0
__vm_munmap+0x75/0x120
__x64_sys_munmap+0x28/0x40
do_syscall_64+0x38/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
...
For each munmap() call, the Xen hypervisor (if built with CONFIG_DEBUG)
would print out the following and trigger a general protection fault in
the affected Xen PV domain:
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
(XEN) d0v... Attempt to implicitly unmap d0's grant PTE ...
As of this writing, gntdev_grant_map structure's vma field (referred to
as map->vma below) is mainly used for checking the start and end
addresses of mappings. However, with split VMAs, these may change, and
there could be more than one VMA associated with a gntdev mapping.
Hence, remove the use of map->vma and rely on map->pages_vm_start for
the original start address and on (map->count << PAGE_SHIFT) for the
original mapping size. Let the invalidate() and find_special_page()
hooks use these.
Also, given that there can be multiple VMAs associated with a gntdev
mapping, move the "mmu_interval_notifier_remove(&map->notifier)" call to
the end of gntdev_put_map, so that the MMU notifier is only removed
after the closing of the last remaining VMA.
Finally, use an atomic to prevent inadvertent gntdev mapping re-use,
instead of using the map->live_grants atomic counter and/or the map->vma
pointer (the latter of which is now removed). This prevents the
userspace from mmap()'ing (with MAP_FIXED) a gntdev mapping over the
same address range as a previously set up gntdev mapping. This scenario
can be summarized with the following call-trace, which was valid prior
to this commit:
mmap
gntdev_mmap
mmap (repeat mmap with MAP_FIXED over the same address range)
gntdev_invalidate
unmap_grant_pages (sets 'being_removed' entries to true)
gnttab_unmap_refs_async
unmap_single_vma
gntdev_mmap (maps the shared pages again)
munmap
gntdev_invalidate
unmap_grant_pages
(no-op because 'being_removed' entries are true)
unmap_single_vma (For PV domains, Xen reports that a granted page
is being unmapped and triggers a general protection fault in the
affected domain, if Xen was built with CONFIG_DEBUG)
The fix for this last scenario could be worth its own commit, but we
opted for a single commit, because removing the gntdev_grant_map
structure's vma field requires guarding the entry to gntdev_mmap(), and
the live_grants atomic counter is not sufficient on its own to prevent
the mmap() over a pre-existing mapping.
Link: https://github.com/QubesOS/qubes-issues/issues/7631
Fixes: ab31523c2fca ("xen/gntdev: allow usermode to map granted pages")
Cc: stable@vger.kernel.org
Signed-off-by: M. Vefa Bicakci <m.v.b@runbox.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/20221002222006.2077-3-m.v.b@runbox.com
Signed-off-by: Juergen Gross <jgross@suse.com>
2022-10-02 18:20:06 -04:00
|
|
|
|
|
|
|
|
|
2019-11-12 16:22:31 -04:00
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2011-01-07 11:51:47 +00:00
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
2020-01-28 10:31:26 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-06-08 21:33:54 -07:00
|
|
|
|
2020-01-28 10:31:26 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-06-28 03:21:41 -07:00
|
|
|
|
2011-02-03 14:16:54 -05:00
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
2018-07-20 12:01:47 +03:00
|
|
|
|
2011-02-03 14:16:54 -05:00
|
|
|
|
|
|
|
|
|
2011-01-07 11:51:47 +00:00
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
xen/gntdev.c: Replace vm_map_pages() with vm_map_pages_zero()
'commit df9bde015a72 ("xen/gntdev.c: convert to use vm_map_pages()")'
breaks gntdev driver. If vma->vm_pgoff > 0, vm_map_pages()
will:
- use map->pages starting at vma->vm_pgoff instead of 0
- verify map->count against vma_pages()+vma->vm_pgoff instead of just
vma_pages().
In practice, this breaks using a single gntdev FD for mapping multiple
grants.
relevant strace output:
[pid 857] ioctl(7, IOCTL_GNTDEV_MAP_GRANT_REF, 0x7ffd3407b6d0) = 0
[pid 857] mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_SHARED, 7, 0) =
0x777f1211b000
[pid 857] ioctl(7, IOCTL_GNTDEV_SET_UNMAP_NOTIFY, 0x7ffd3407b710) = 0
[pid 857] ioctl(7, IOCTL_GNTDEV_MAP_GRANT_REF, 0x7ffd3407b6d0) = 0
[pid 857] mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_SHARED, 7,
0x1000) = -1 ENXIO (No such device or address)
details here:
https://github.com/QubesOS/qubes-issues/issues/5199
The reason is -> ( copying Marek's word from discussion)
vma->vm_pgoff is used as index passed to gntdev_find_map_index. It's
basically using this parameter for "which grant reference to map".
map struct returned by gntdev_find_map_index() describes just the pages
to be mapped. Specifically map->pages[0] should be mapped at
vma->vm_start, not vma->vm_start+vma->vm_pgoff*PAGE_SIZE.
When trying to map grant with index (aka vma->vm_pgoff) > 1,
__vm_map_pages() will refuse to map it because it will expect map->count
to be at least vma_pages(vma)+vma->vm_pgoff, while it is exactly
vma_pages(vma).
Converting vm_map_pages() to use vm_map_pages_zero() will fix the
problem.
Marek has tested and confirmed the same.
Cc: stable@vger.kernel.org # v5.2+
Fixes: df9bde015a72 ("xen/gntdev.c: convert to use vm_map_pages()")
Reported-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Signed-off-by: Souptick Joarder <jrdr.linux@gmail.com>
Tested-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
2019-07-31 00:04:56 +05:30
|
|
|
|
2019-05-13 17:22:23 -07:00
|
|
|
|
|
|
|
|
|
2011-02-03 12:19:02 -05:00
|
|
|
|
|
|
|
|
|
2011-01-07 11:51:47 +00:00
|
|
|
|
|
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
2011-02-03 14:16:54 -05:00
|
|
|
|
2011-03-19 08:45:43 +03:00
|
|
|
|
2015-01-09 18:06:12 +00:00
|
|
|
|
2011-02-03 14:16:54 -05:00
|
|
|
|
2025-08-26 16:56:08 +02:00
|
|
|
|
2018-01-09 12:10:22 +00:00
|
|
|
|
2013-01-02 22:57:12 +00:00
|
|
|
|
2011-02-03 14:16:54 -05:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-06-28 03:21:41 -07:00
|
|
|
|
2010-12-14 18:40:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
