ACPI: bus.c, fix error handling in acpi_bus_init
PATCH 1/1 ACPI: bus.c, fix error handling in acpi_bus_init
There was a misplaced status test. Move it to correct place and rollback appropriately.
--- drivers/acpi/bus.c | 12 ++++++------ 1 files changed, 6 insertions(+), 6 deletions(-)
drivers/acpi/bus.c +6 -6
@@ -760,18 +760,17 @@ static int __init acpi_bus_init(void) status =acpi_os_initialize1 (); - - status = -acpi_enable_subsystem (ACPI_NO_HARDWARE_INIT |ACPI_NO_ACPI_ENABLE );if (ACPI_FAILURE (status)) { printk(KERN_ERR PREFIX - "Unable to start the ACPI Interpreter\n"); - goto error1; + "Unable to initialize ACPI OS objects\n"); + goto error0; } + status = +acpi_enable_subsystem (ACPI_NO_HARDWARE_INIT |ACPI_NO_ACPI_ENABLE );if (ACPI_FAILURE (status)) { printk(KERN_ERR PREFIX - "Unable to initialize ACPI OS objects\n"); + "Unable to start the ACPI Interpreter\n"); goto error1; }
@@ -832,6 +831,7 @@ static int __init acpi_bus_init(void)/* Mimic structured exception handling */ error1:acpi_terminate (); +error0:return -ENODEV ; }
Signature · 1 line
1.6.1.3
Len Brown
reviewerQuoted 2 lines · Jiri Slaby
There was a misplaced status test. Move it to correct place and rollback appropriately.
hmm, looks like this has been broken since the day acpi_os_initialize1() was invented in 2004.
Turns out that the bug and its fix are moot, however, as acpi_os_initialize1() is hard-coded to return success, opting for BUG_ON() if it sees a failure...
So unless that changes, I'd prefer to keep the code simple and just not check its status -- which is effectively what we're doing right now.
thanks,
-Len
Quoted 44 lines
Signed-off-by Jiri Slaby <[email protected]>--- drivers/acpi/bus.c | 12 ++++++------ 1 files changed, 6 insertions(+), 6 deletions(-)drivers/acpi/bus.c +6
-6@@ -760,18 +760,17 @@ static int __init acpi_bus_init(void) status =acpi_os_initialize1 (); - - status = -acpi_enable_subsystem (ACPI_NO_HARDWARE_INIT |ACPI_NO_ACPI_ENABLE );if (ACPI_FAILURE (status)) { printk(KERN_ERR PREFIX - "Unable to start the ACPI Interpreter\n"); - goto error1; + "Unable to initialize ACPI OS objects\n"); + goto error0; } + status = +acpi_enable_subsystem (ACPI_NO_HARDWARE_INIT |ACPI_NO_ACPI_ENABLE );if (ACPI_FAILURE (status)) { printk(KERN_ERR PREFIX - "Unable to initialize ACPI OS objects\n"); + "Unable to start the ACPI Interpreter\n"); goto error1; }@@ -832,6 +831,7 @@ static int __init acpi_bus_init(void)/* Mimic structured exception handling */ error1:acpi_terminate (); +error0:return -ENODEV ; }Signature · 1 line
1.6.1.3
Quoted 7 lines · Len Brown
Quoted 2 lines · Jiri Slaby
There was a misplaced status test. Move it to correct place and rollback appropriately.
hmm, looks like this has been broken since the day
acpi_os_initialize1()was invented in 2004.
It comes from bitkeeper times, I checked, it looked like a typo.
Quoted 3 lines
Turns out that the bug and its fix are moot, however, as
acpi_os_initialize1()is hard-coded to return success, opting forBUG_ON()if it sees a failure...
BUG_ON is not noreturn e.g. on embedded, anyway it would crash later in this case. I was thinking about returning a failure when one of them is NULL, but fell back to not do so and check only retval in the caller.
Quoted 3 lines
So unless that changes, I'd prefer to keep the code simple and just not check its status -- which is effectively what we're doing right now.
Ok, no problem. Will repost.
Thanks.