I'm a bit confused. What does the execution of the _Sx methods have to do with module-level code? Obviously, many times the _Sx objects are implemented as simply named objects, but the table load doesn't care what is and what isn't present. > -----Original Message----- > From: Zheng, Lv > Sent: Friday, October 09, 2015 6:38 PM > To: Prarit Bhargava; Moore, Robert; devel@xxxxxxxxxx > Cc: linux-acpi@xxxxxxxxxxxxxxx; Wysocki, Rafael J > Subject: RE: [PATCH] ACPICA: AcpiGetSleepTypeData: Failure to find \_Sx > should not result in a loud warning [v2] > > Hi, > > > From: Prarit Bhargava [mailto:prarit@xxxxxxxxxx] > > Sent: Friday, October 09, 2015 7:30 PM > > > > On 10/09/2015 01:26 AM, Zheng, Lv wrote: > > > Please ignore this. > > > The fix is against the caller. > > > > > > Thanks and best regards > > > -Lv > > > > > >> From: Devel [mailto:devel-bounces@xxxxxxxxxx] On Behalf Of Zheng, > > >> Lv > > >> Sent: Friday, October 09, 2015 10:02 AM > > >> > > >> Why don't you fix this in the invoker side? > > >> For example: > > >> If (acpi_get_handle()) > > >> acpi_evaluate_object() > > > > That seems like a sloppy workaround an actual bug in ACPICA. > > So let me ignore this which is a reply for a useless question. > > > > > >> So that the AE_NOT_FOUND warning can still be kept for the real > troubles? > > > > The code is warning 100% of the time on something that is optional. > > Let me ignore this which is a reply for a useless question. > > > > > >> There are really scenarios that such warning is useful for catching > bugs. > > >> > > > > What scenario is possible where this causes a problem? Issuing an > > error on something that is optional is not a good idea. > > There are several scenarios, let me say some known stuffs about the module > level code support. > In AML grammar, module level code (non-object-definition AML opcodes > outside of a control method) is legal. > While it is supported in ACPICA in a split way: > 1. In the early stage, table loading code only executes those object > definition opcodes and links all "if/else/while" AML fragments together. > 2. In the late stage, object initialization code will execute such AML > fragments before invoking all _INI methods. > > 1. Since code is not executed in the Windows compliant order during the > early table loading stage, some objects are not defined while they should. > 2. Even worse, during the early stage, ACPICA may fail the whole table > loading process due to AE_NOT_FOUND, leaving us partial initialized > objects (the ACPICA AML interpreter implements 2 phases to parse a table > to complete the table loading). > Such issue is worked around by deleting the namespace objects created > during the early stage if an exception (AE_NOT_FOUND) is encountered. > > So you can see, such kind of error is sometime useful for catching AML > interpreter issues or BIOS bugs. > > Thanks and best regards > -Lv -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
