[PATCH 0/2] dynamic table unload fixes

[PATCH 0/2] dynamic table unload fixes

[Bob Paauwe]

While trying to make use of dynamic tables as a way to provide driver
initialization properties, I've encountered a couple of issues with the
way dynamic tables are being handled.

The first patch appears to be a bug. There is a mutex release without
a corresponding aquire. This extra release causes a mutex error message
when a table is re-loaded after being unloaded.

The second patch actually unloads a table when table unload is called. 
When a table is being dynamically loaded by a driver module and that
driver module is unloaded, the memory for the table would typically be
freed resulting in a table that is marked as unloaded but still pointing
to that freed memory.  If the driver module is then re-loaded, the
ACPI table reload function will attempt to access that freed memory when
it does a table compare and results in a oops.

These patches solve the problems I've encountered but may not be the 
proper way to solve.

Bob Paauwe (2):
  Remove extra release mutex call.
  Acutally unload a table.

 drivers/acpi/acpica/tbinstal.c | 1 -
 drivers/acpi/acpica/tbxfload.c | 1 +
 2 files changed, 1 insertion(+), 1 deletion(-)

-- 
2.1.0

[PATCH 1/2] Remove extra release mutex call.

[Bob Paauwe]

Remove the release mutex call when a table is re-loaded. This seems to
be here to release a mutex that was aquired at one time in table verify
function.  That aquire is no longer present, so remove the release.

Signed-off-by: Bob Paauwe <bob.j.paauwe@intel.com>
---
 drivers/acpi/acpica/tbinstal.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/drivers/acpi/acpica/tbinstal.c b/drivers/acpi/acpica/tbinstal.c
index 9bad45e..7fbc2b9 100644
--- a/drivers/acpi/acpica/tbinstal.c
+++ b/drivers/acpi/acpica/tbinstal.c
@@ -346,7 +346,6 @@ acpi_tb_install_standard_table(acpi_physical_address address,
 				 */
 				acpi_tb_uninstall_table(&new_table_desc);
 				*table_index = i;
-				(void)acpi_ut_release_mutex(ACPI_MTX_TABLES);
 				return_ACPI_STATUS(AE_OK);
 			}
 		}
-- 
2.1.0

[PATCH 2/2] Acutally unload a table.

[Bob Paauwe]

Instead of simply marking it as unloaded.  If a table is dynamically
loaded by a module and the module is then unloaded, the table is no
longer valid so it's best to make sure it is unloaded.

Signed-off-by: Bob Paauwe <bob.j.paauwe@intel.com>
---
 drivers/acpi/acpica/tbxfload.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/acpi/acpica/tbxfload.c b/drivers/acpi/acpica/tbxfload.c
index aadb300..ce75ca2 100644
--- a/drivers/acpi/acpica/tbxfload.c
+++ b/drivers/acpi/acpica/tbxfload.c
@@ -408,6 +408,7 @@ acpi_status acpi_unload_parent_table(acpi_handle object)
 
 		status = acpi_tb_release_owner_id(i);
 		acpi_tb_set_table_loaded_flag(i, FALSE);
+		acpi_tb_uninstall_table(&acpi_gbl_root_table_list.tables[i]);
 		break;
 	}
 
-- 
2.1.0

Re: [PATCH 0/2] dynamic table unload fixes

[Rafael J. Wysocki]

On Friday, April 03, 2015 09:07:47 AM Bob Paauwe wrote:
> While trying to make use of dynamic tables as a way to provide driver
> initialization properties, I've encountered a couple of issues with the
> way dynamic tables are being handled.
> 
> The first patch appears to be a bug. There is a mutex release without
> a corresponding aquire. This extra release causes a mutex error message
> when a table is re-loaded after being unloaded.
> 
> The second patch actually unloads a table when table unload is called. 
> When a table is being dynamically loaded by a driver module and that
> driver module is unloaded, the memory for the table would typically be
> freed resulting in a table that is marked as unloaded but still pointing
> to that freed memory.  If the driver module is then re-loaded, the
> ACPI table reload function will attempt to access that freed memory when
> it does a table compare and results in a oops.
> 
> These patches solve the problems I've encountered but may not be the 
> proper way to solve.
> 
> Bob Paauwe (2):
>   Remove extra release mutex call.
>   Acutally unload a table.
> 
>  drivers/acpi/acpica/tbinstal.c | 1 -
>  drivers/acpi/acpica/tbxfload.c | 1 +
>  2 files changed, 1 insertion(+), 1 deletion(-)

This is ACPICA material.  Please *at* *least* CC it to the ACPICA maintainers
as per MAINTAINERS, but it really should be sent to the ACPICA list (and you
can CC linux-acpi if that is useful too).


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.