ACPI issues - suspend-to-ram and processor speed

ACPI issues - suspend-to-ram and processor speed

[Andre Eisenbach]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dear ACPI devel team,

I have two issues which prevent me from fully enjoying my notbook running on 
Linux. Both are seemingly related to ACPI. Please help me understand the 
problems better and maybe fix them or help me attempting a fix.

The notebook in question is a Compaq Presario 2100Z with an AMD Athlon XP 
2400+ processor. Current kernel version is 2.6.6-mm2. I am not applying any 
ACPI patches from sf.net, but rely on whats in the kernel (which may be my 
first mistake).

I am very willing to learn and may be able to fix it myself, if you guys can 
help me find out more about which components are involved.

Problem 1 - Cannot suspend (or resume better)

Right now I am unable to suspend my notebook in any for or fashion (ram, disk, 
etc.). Therefore I have to turn off the notebook and boot again, which 
honestly makes the notebook pretty useless.

When I try to suspend to RAM, the suspend part seems to run fine (I can see 
suspend-to-ram messages), the lcd shuts off and the power button starts 
pulsing slowly (indicating standby). So far so good. However, if I hit the 
power button then to resume, the device seems to want to wake up (hd spins 
up, fans start) but the LCD never turns on and the notebook quickly (1-2 
seconds) later switched to off (all quiet/lights off). Hitting the power 
button again starts a normal startup.

I also tried to suspend to disk using swsusp. That corrupted my swap partition 
and didn't resume.

What does it take to suspend to RAM?
What could be broken, and where should I start looking if I want to fix it?
What more debug information do you guys need?

Problem 2 - CPU slows down when on battery

When the power cord is unplugged, my CPU slows down quite a bit. This is not 
indicated in /proc/cpuinfo and NOT due to cpufreq / powernow adjustments.

The problem does not seem to occur with ACPI=OFF 

More details here in a post to the linux-kernel mailing list:
http://lkml.org/lkml/2004/5/9/100

Also a relevant and informative entry here:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=467655

When on battery, KDE is barely usable. Once again it doesnt make much sense 
for a laptop not to be usable on battery ... :(


Thanks for your attention and hopefully for your answers/comments. I can 
provide as much debug/info/configs as you would like and as a programmer am 
willing to help out/read up, if you just help me identify where to start.

Thanks,
   André Eisenbach
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFAqdO45DBE808V5icRAnEcAJ49pvxNlggXYpLBARhqh/pNSmSKewCgr6Eg
mW5uBTiyk0Ab2IdtDXV3038=
=lqX+
-----END PGP SIGNATURE-----


-------------------------------------------------------
This SF.Net email is sponsored by: SourceForge.net Broadband
Sign-up now for SourceForge Broadband and get the fastest
6.0/768 connection for only $19.95/mo for the first 3 months!
http://ads.osdn.com/?ad_id%62&alloc_ida84&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Mathieu LESNIAK]

Hello,

I confirm what Andre have said : I've got a Presario 2118EA (quite same 
model with AMD Athlon XP1800). Suspend to ram hav never worked for me. 
It suspends but cannot resume (2 seconds after the wake up, the laptop 
shuts down)
I suspected acpid to catch the power button press and tell the laptop to 
shut down, but after disabling it, no more luck...

Regarding the CPU slowdown, I've seen similar behaviour under Windows. I 
think it's related to the laptop itself, maybe as Pavel Machek said on 
LKML, it's the lack of power that force the laptop to run in this mode.

Mathieu LESNIAK

Andre Eisenbach wrote:
> Dear ACPI devel team,
> 
> I have two issues which prevent me from fully enjoying my notbook running on
> Linux. Both are seemingly related to ACPI. Please help me understand the
> problems better and maybe fix them or help me attempting a fix.
> 
> The notebook in question is a Compaq Presario 2100Z with an AMD Athlon XP
> 2400+ processor. Current kernel version is 2.6.6-mm2. I am not applying any
> ACPI patches from sf.net, but rely on whats in the kernel (which may be my
> first mistake).
> 
> I am very willing to learn and may be able to fix it myself, if you guys can
> help me find out more about which components are involved.
> 
> Problem 1 - Cannot suspend (or resume better)
> 
> Right now I am unable to suspend my notebook in any for or fashion (ram, disk,
> etc.). Therefore I have to turn off the notebook and boot again, which
> honestly makes the notebook pretty useless.
> 
> When I try to suspend to RAM, the suspend part seems to run fine (I can see
> suspend-to-ram messages), the lcd shuts off and the power button starts
> pulsing slowly (indicating standby). So far so good. However, if I hit the
> power button then to resume, the device seems to want to wake up (hd spins
> up, fans start) but the LCD never turns on and the notebook quickly (1-2
> seconds) later switched to off (all quiet/lights off). Hitting the power
> button again starts a normal startup.
> 
> I also tried to suspend to disk using swsusp. That corrupted my swap partition
> and didn't resume.
> 
> What does it take to suspend to RAM?
> What could be broken, and where should I start looking if I want to fix it?
> What more debug information do you guys need?
> 
> Problem 2 - CPU slows down when on battery
> 
> When the power cord is unplugged, my CPU slows down quite a bit. This is not
> indicated in /proc/cpuinfo and NOT due to cpufreq / powernow adjustments.
> 
> The problem does not seem to occur with ACPI=OFF
> 
> More details here in a post to the linux-kernel mailing list:
> http://lkml.org/lkml/2004/5/9/100
> 
> Also a relevant and informative entry here:
> http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=467655
> 
> When on battery, KDE is barely usable. Once again it doesnt make much sense
> for a laptop not to be usable on battery ... :(
> 
> 
> Thanks for your attention and hopefully for your answers/comments. I can
> provide as much debug/info/configs as you would like and as a programmer am
> willing to help out/read up, if you just help me identify where to start.
> 
> Thanks,
>    André Eisenbach

-------------------------------------------------------
This SF.Net email is sponsored by: SourceForge.net Broadband
Sign-up now for SourceForge Broadband and get the fastest
6.0/768 connection for only $19.95/mo for the first 3 months!
http://ads.osdn.com/?ad_id%62&alloc_ida84&op=click
_______________________________________________
Acpi-devel mailing list
Acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
https://lists.sourceforge.net/lists/listinfo/acpi-devel




-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g.
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_id66&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Andre Eisenbach]

Mathieu LESNIAK wrote:

> Regarding the CPU slowdown, I've seen similar behaviour under Windows. 
> I think it's related to the laptop itself, maybe as Pavel Machek said 
> on LKML, it's the lack of power that force the laptop to run in this 
> mode.


According to the post on the LKML, the slowdown dows not occur in Linux 
with ACPI=off.
I do think that it may be a BIOS problem, but I hope that the ACPI 
knowledgable may be able to provide a workaround.

Cheers,
    André


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g.
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_id66&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Len Brown]

On Thu, 2004-05-20 at 13:52, Andre Eisenbach wrote:
> Mathieu LESNIAK wrote:
> 
> > Regarding the CPU slowdown, I've seen similar behaviour under Windows. 
> > I think it's related to the laptop itself, maybe as Pavel Machek said 
> > on LKML, it's the lack of power that force the laptop to run in this 
> > mode.
> 
> 
> According to the post on the LKML, the slowdown dows not occur in Linux 
> with ACPI=off.
> I do think that it may be a BIOS problem, but I hope that the ACPI 
> knowledgable may be able to provide a workaround.

could be a number of things.  Disabling the processor module would be
the workaround.

-Len




-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Pavel Machek]

Hi!

> I also tried to suspend to disk using swsusp. That corrupted my swap partition 
> and didn't resume.

Read swsusp.txt, power/tricks.txt.


-- 
64 bytes from 195.113.31.123: icmp_seq=28 ttl=51 time=448769.1 ms         



-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Andre Eisenbach]

[-- Attachment #1: Type: text/plain, Size: 826 bytes --]

Am Sonntag Mai 23 2004 05:29 schrieb Pavel Machek:
> > I also tried to suspend to disk using swsusp. That corrupted my swap
> > partition and didn't resume.
>
> Read swsusp.txt, power/tricks.txt.

Thank you, thank you, thank you.
RTFM seems to have done the trick.

After reading the two files suggested, I recompiled my kernel (2.6.7-rc1-mm1) 
with CONFIG_PM_DISK and _without_ CONFIG_PREEMPT and for the first time ever 
got software suspend to work!

It takes about 50secs to resume, which is still a relatively long time, but 
its definitely less than the about 1:20min it takes for me to fully boot and 
log on AND my state is preserved.

My notebook just got a whole lot more usable!
Thanks ACPI devs!

Now, if only suspend-to-ram would work.

How could I debug this problem?

Thanks,
   André

[-- Attachment #2: signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: ACPI issues - suspend-to-ram and processor speed

[Pavel Machek]

Hi!

> > > I also tried to suspend to disk using swsusp. That corrupted my swap
> > > partition and didn't resume.
> >
> > Read swsusp.txt, power/tricks.txt.
> 
> Thank you, thank you, thank you.
> RTFM seems to have done the trick.
> 
> After reading the two files suggested, I recompiled my kernel (2.6.7-rc1-mm1) 
> with CONFIG_PM_DISK and _without_ CONFIG_PREEMPT and for the first time ever 
> got software suspend to work!

Actually, stay away from CONFIG_PM_DISK, CONFIG_SOFTWARE_SUSPEND
should work better.


> My notebook just got a whole lot more usable!
> Thanks ACPI devs!
> 
> Now, if only suspend-to-ram would work.
> 
> How could I debug this problem?

Well, it involves assembly, black magic and soldering iron. Still
interested?
								Pavel
-- 
934a471f20d6580d5aad759bf0d97ddc


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Andre Eisenbach]

[-- Attachment #1: Type: text/plain, Size: 844 bytes --]

Am Freitag Mai 28 2004 01:24 schrieb Pavel Machek:
> > After reading the two files suggested, I recompiled my kernel
> > (2.6.7-rc1-mm1) with CONFIG_PM_DISK and _without_ CONFIG_PREEMPT and for
> > the first time ever got software suspend to work!
>
> Actually, stay away from CONFIG_PM_DISK, CONFIG_SOFTWARE_SUSPEND
> should work better.

Ok, recompiled with CONFIG_SOFTWARE_SUSPEND and it still works :)

> > How could I debug this problem?
>
> Well, it involves assembly, black magic and soldering iron. Still
> interested?

Well, I am a programmer - albeit with rusty assembly knowledge, I do own a 
soldering iron and I own a black cat. Does that count? ;)

Where can I start to learn more about ACPI and which pieces have to be in 
place for suspend-to-ram to work?

Thanks for all your help!!

Cheers,
    André

[-- Attachment #2: signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: ACPI issues - suspend-to-ram and processor speed

[Pavel Machek]

Hi!

> > > How could I debug this problem?
> >
> > Well, it involves assembly, black magic and soldering iron. Still
> > interested?
> 
> Well, I am a programmer - albeit with rusty assembly knowledge, I do own a 
> soldering iron and I own a black cat. Does that count? ;)
> 
> Where can I start to learn more about ACPI and which pieces have to be in 
> place for suspend-to-ram to work?

Take a look at arch/i386/kernel/acpi/wakeup.S, there's where resume
begins. Try to find out if wakeup_code: is executed or not. (That's
where soldering iron comes to play). If wakeup_code is not executed,
you'll need to fix sleep code. If it is, wakeup needs to be fixed...

								Pavel
-- 
934a471f20d6580d5aad759bf0d97ddc


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Andre Eisenbach]

[-- Attachment #1: Type: text/plain, Size: 1230 bytes --]

Stupid question of the evening:

It's 5am here and I've just spent about the whole night playing witch ACPI, 
DSDTs etc...

The notebook I have can only resume from standby by pressing the power button. 
Could it be, that when the power button is pressed, the system starts to wake 
up, but the "power button event" is not intercepted by ACPI and passes 
through to "the hardware", causing the notebook to shut down for good?

I know thats in very rough laymans terms, but I hope it's enough for you guys 
to tell me just how wrong I am ;).

Here's a rought flowchart of what's happening on my notebook (afer 6. is the 
problem):

1. echo 3 > proc/acpi/sleep
2. PM: Preparing system for suspend
3. Stopping tasks: ==================|

4. --> Notebook enters standby. HD quiet, LCD off, fans off, On/Off LED pulses 
slowly

5. Press power button to wake notebok

6. --> Notebook tries to leave standby. HD spins up, LCD light on but no text, 
fans start...
7. --> Not even 1/2 second later, all off. Notebook is off. On/Off LED off

8. Press power button again.
9. --> Normal boot...

Could it be that there's some magic power-off event queued up when you press 
the button?

Thanks,
    André

[-- Attachment #2: signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: ACPI issues - suspend-to-ram and processor speed

[Pavel Machek]

Hi!

Can you get any data out of notebook? Like 8 leds on paralel port, pci
card decoding port 0x80 etc? [That's where soldering iron might be usefull].

> Stupid question of the evening:
> 
> It's 5am here and I've just spent about the whole night playing witch ACPI, 
> DSDTs etc...
> 
> The notebook I have can only resume from standby by pressing the power button. 
> Could it be, that when the power button is pressed, the system starts to wake 
> up, but the "power button event" is not intercepted by ACPI and passes 
> through to "the hardware", causing the notebook to shut down for
> good?

Improbable.

> I know thats in very rough laymans terms, but I hope it's enough for you guys 
> to tell me just how wrong I am ;).
> 
> Here's a rought flowchart of what's happening on my notebook (afer 6. is the 
> problem):
> 
> 1. echo 3 > proc/acpi/sleep
> 2. PM: Preparing system for suspend
> 3. Stopping tasks: ==================|
> 
> 4. --> Notebook enters standby. HD quiet, LCD off, fans off, On/Off LED pulses 
> slowly
> 
> 5. Press power button to wake notebok
> 
> 6. --> Notebook tries to leave standby. HD spins up, LCD light on but no text, 
> fans start...
> 7. --> Not even 1/2 second later, all off. Notebook is off. On/Off
> LED off

Easy. Insert infinite loop at begining of wakeup_start. If it hangs,
you know it reached wakeup_start, and fix wakeup code. If it does not,
we left hardware at "bad" (i.e. bios did not like it) state before
suspend, fix suspend code. (Turn off APIC etc). [This is where your
assembly skills come handy.]

> 8. Press power button again.
> 9. --> Normal boot...
> 
> Could it be that there's some magic power-off event queued up when you press 
> the button?

If ACPI event was queued, it would do clean shutdown; you'd notice
that.
								Pavel
PS: When stuck, try asking your black cat :-).

-- 
934a471f20d6580d5aad759bf0d97ddc


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Andre Eisenbach]

[-- Attachment #1: Type: text/plain, Size: 988 bytes --]

Am Samstag Mai 29 2004 04:45 schrieb Pavel Machek:
> Can you get any data out of notebook? Like 8 leds on paralel port, pci
> card decoding port 0x80 etc? [That's where soldering iron might be
> usefull].

Yes, I can hook up the parallel port. Just let me know where we're gonna use 
it.

> > 7. --> Not even 1/2 second later, all off. Notebook is off. On/Off
> > LED off
>
> Easy. Insert infinite loop at begining of wakeup_start. If it hangs,
> you know it reached wakeup_start, and fix wakeup code. If it does not,
> we left hardware at "bad" (i.e. bios did not like it) state before
> suspend, fix suspend code. (Turn off APIC etc). [This is where your
> assembly skills come handy.]

I put an endless loop into wakeup_start. Nothing changed. Same behaviour. I 
also recompiled the kernel without APIC, but I'm not sure thats what you 
meant.

Where do I find the suspend code?

> PS: When stuck, try asking your black cat :-).

She's sleeping. A lot.

;)

[-- Attachment #2: signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: ACPI issues - suspend-to-ram and processor speed

[Pavel Machek]

Hi!

> Am Samstag Mai 29 2004 04:45 schrieb Pavel Machek:
> > Can you get any data out of notebook? Like 8 leds on paralel port, pci
> > card decoding port 0x80 etc? [That's where soldering iron might be
> > usefull].
> 
> Yes, I can hook up the parallel port. Just let me know where we're gonna use 
> it.

Not yet, but something like that is neccessary for debugging wakeup
phase.

> > > 7. --> Not even 1/2 second later, all off. Notebook is off. On/Off
> > > LED off
> >
> > Easy. Insert infinite loop at begining of wakeup_start. If it hangs,
> > you know it reached wakeup_start, and fix wakeup code. If it does not,
> > we left hardware at "bad" (i.e. bios did not like it) state before
> > suspend, fix suspend code. (Turn off APIC etc). [This is where your
> > assembly skills come handy.]
> 
> I put an endless loop into wakeup_start. Nothing changed. Same behaviour. I 
> also recompiled the kernel without APIC, but I'm not sure thats what you 
> meant.

Okay, so problem is with suspend. Recompiling without APIC was good
idea, see also tricks.txt what more you can do.

> Where do I find the suspend code?

Its all around acpi, last thing it does is in
drivers/acpi/hardware/hwsleep.c.
									Pavel
-- 
934a471f20d6580d5aad759bf0d97ddc


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Vibol Hou]

Where should one look if system doesn't do the slowly blinking power LED 
when echo 3 > /proc/acpi/sleep?  It says entering state and then appears 
to just power off, nothing flashing.

-Vibol

>Take a look at arch/i386/kernel/acpi/wakeup.S, there's where resume
>begins. Try to find out if wakeup_code: is executed or not. (That's
>where soldering iron comes to play). If wakeup_code is not executed,
>you'll need to fix sleep code. If it is, wakeup needs to be fixed...
>
>								Pavel
>  
>



-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Pavel Machek]

Hi!


> Where should one look if system doesn't do the slowly blinking power LED 
> when echo 3 > /proc/acpi/sleep?  It says entering state and then appears 
> to just power off, nothing flashing.

ACPI problem while entering suspend... Systems were where writing
wrong number to 4-bit field somewhere in hwsleep.c actually made it to
work...

								Pavel

> -Vibol
> 
> >Take a look at arch/i386/kernel/acpi/wakeup.S, there's where resume
> >begins. Try to find out if wakeup_code: is executed or not. (That's
> >where soldering iron comes to play). If wakeup_code is not executed,
> >you'll need to fix sleep code. If it is, wakeup needs to be fixed...
> >
> >								Pavel
> > 
> >
> 

-- 
934a471f20d6580d5aad759bf0d97ddc


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Cam]

Pavel

> ACPI problem while entering suspend... Systems were where writing
> wrong number to 4-bit field somewhere in hwsleep.c actually made it to
> work...

When my system suspends/resumes, which is a very risk process with 
recent kernels (for many builds it hangs on resume), there is a text 
console that appears.

Sometimes I get 4 characters corruption on that console in a strange 
colour... could that be related?

-Cam


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Pavel Machek]

Hi!

> >ACPI problem while entering suspend... Systems were where writing
> >wrong number to 4-bit field somewhere in hwsleep.c actually made it to
> >work...
> 
> When my system suspends/resumes, which is a very risk process with 
> recent kernels (for many builds it hangs on resume), there is a text 
> console that appears.

So new kernels have it broken but it worked in old ones?

> Sometimes I get 4 characters corruption on that console in a strange 
> colour... could that be related?

Is it trying to write "Linu" in yellow-on-black? That is my debugging
hack, look i386/kernel/acpi/wakeup.S.
								Pavel
-- 
934a471f20d6580d5aad759bf0d97ddc


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click

Re: ACPI issues - suspend-to-ram and processor speed

[Pavel Machek]

Hi!

> Where should one look if system doesn't do the slowly blinking power LED 
> when echo 3 > /proc/acpi/sleep?  It says entering state and then appears 
> to just power off, nothing flashing.

hwsleep.c
								Pavel

-- 
934a471f20d6580d5aad759bf0d97ddc


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click