From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ducrot Bruno Subject: Re: Compaq Evo N610c in almost perfect ACPI support shocker Date: Wed, 1 Oct 2003 19:42:32 +0200 Sender: acpi-devel-admin-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org Message-ID: <20031001174232.GG11391@poupinou.org> References: <20031001163539.GA22358@clara.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20031001163539.GA22358-6JSjyQ0Qj1ReoWH0uzbU5w@public.gmane.org> Errors-To: acpi-devel-admin-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org List-Help: List-Post: List-Subscribe: , List-Unsubscribe: , List-Archive: To: Faye Pearson Cc: acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org List-Id: linux-acpi@vger.kernel.org On Wed, Oct 01, 2003 at 05:35:39PM +0100, Faye Pearson wrote: > I've just received a Compaq Evo N610c which I wanted to put Linux on. > I'd previously owned a Presario 2701ea which had a buggy DSDT (usual off > by one errors etc) so was expecting a bit of a bumpy ride. > > Kudos to HP/Compaq though, the new BIOS runs almost flawlessly through > Linux's ACPI and the only errors on boot seem to be on the interpreter > side (although I'd love to be proved wrong ;) > > One of the problems seems to be long standing - I found a kludge on the > mailing list to fix it: > > Method (C051, 4, NotSerialized) > { > // Copy Arg1 pointer to Local2 (in theory) > Store (Arg1, Local2) > Store (Arg3, Local4) > Add (Local4, 0x1A, Local5) > While (LLess (Local4, Local5)) > { > Store (DerefOf (Index (C04F, Local4)), Index (Arg0, > Local2)) > Increment (Local4) > // Nasty side-effect, also increments Arg1 > Increment (Local2) > } > > // Let's reset Local2 to Arg1 again.. oh dear it's the same value. > // Store (Arg1, Local2) > // Here comes the kludge > Subtract (Local2, 0x1A, Local2) > // This was making it go past end of buffer > Add (Local2, 0x0A, Local2) > > As you can see I've applied the Subtract kludge other people have used, > but there's nothing actually wrong with the ASL here AFAICS, it should > work. Ah yes. That seems funny. I wasn't aware for that. > > The other part may just be ACPI being over-verbose on something it's > handling correctly - I don't think I'm losing any functionality because > of it: > > ACPI-1121: *** Error: Method execution failed > [\_SB_.C045.C059.C0E2.C13F] (N > ode cff95e40), AE_AML_UNINITIALIZED_LOCAL > ACPI-1121: *** Error: Method execution failed > [\_SB_.C045.C059.C0E2.C14E] (N > ode cff931c0), AE_AML_UNINITIALIZED_LOCAL > ACPI-1121: *** Error: Method execution failed [\_SB_.C198._BTP] > (Node cff8ec > 80), AE_AML_UNINITIALIZED_LOCAL > ACPI: Battery Slot [C198] (battery present) > > Here's the DSDT snippet: > Method (C13F, 3, Serialized) > { > Add (Arg2, 0x01, Local3) > Store (Buffer (Local3) > { > 0x10 > }, Local2) > Store (0x0A, Local5) > Store (0x01, Local1) > // This next line has been added by me to satisfy the parser > Store (0x00, Local0) > > While (LAnd (Local1, LGreater (Local5, 0x00))) > { > Store (C12E (0x00), Local1) > If (LNot (Local1)) > { > Store (C12D (Arg0, Arg1), Local1) > If (LAnd (LNot (Local1), LGreater (Arg2, > 0x00))) > { > Store (C12C (Arg0, Arg2), Local0) > Store (DerefOf (Index (Local0, > 0x00)), Local > 1) > } > > Or (C12F (), Local1, Local1) > } > > Decrement (Local5) > If (LAnd (Local1, LGreater (Local5, 0x00))) > { > Sleep (0x64) > } > } > > If (Local1) > { > Return (Local2) > } > > // Not initialised if the While loop isn't entered, but if that's the > // case then Local1 is true (1) and this point is not reached. > > Return (Local0) > } > > With the kludge I apply above, there are no errors during boot. I don't have fully looked the while loop, but it seems for me you can have Local0 not initialised, but Local1 contains 0. > Now, onto the HP/Compaq problem I've found (where do I apply for a > ROMPAQ?) There are only two settings for Intel Speedstep in the BIOS > (fine thought I) Auto and Disabled. Disabled, as expected, disables the > Speedstep functionality and you see in > processor...performance. Good so far. > > Auto means that the laptop fully controls the Speedstep functionality > and it cannot be overridden by the OS. I've tested this by installing > CPUFREQ which accesses speedstep outside of ACPI, configuring it for > 60%60%powersave and when running glxgears I get a jump up to the same > FPS I get on 100%100%performance. I can confirm this for several > seconds afterward with cat /proc/sys/cpu/0/speed showing the maximum cpu > speed. ACPI is not being informed of the processor state changes and > shows P0: regardless of which P state the laptop is in. > I don't understand the cpufreq trouble. But my english is bad... What cpufreq driver are you using btw (the ich or the newer smi one?). the smi should work like you want (that is no speedstep if you disable it), but the ich one talk directly to the hardware, regardless of what 'think' bios. And, (but that may change soon), acpi performance and cpufreq have diferent API, so no, if you are using acpi P-state and cpufreq at the same time, then only weird things, confusions about the 2 way happens (that is true *only* for the 2.4 kernel, not for 2.6). -- Ducrot Bruno -- Which is worse: ignorance or apathy? -- Don't know. Don't care. ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf