sesam.hu

Engineering Manager / Trail Runner / Budapest, Hungary

Mountain Lion and the freezes

Not that I expected otherwise, but Mountain Lion did naught in fixing the iMac graphic driver freezes that Apple claimed are the result of hardware issues. If anything, the crashes are more frequent, albeit it can just be caused by the fact that it’s Summer and with no A/C it’s 30°C+ inside.

The freezes are often somewhat more graceful now with a kernel panic message and an automatic restart. After such an event a report is generated and sent to Apple. Not that I have illusions of them taking it seriously.

Interval Since Last Panic Report: 170834 sec
Panics Since Last Report: 3
Anonymous UUID:

Sun Aug 26 15:02:46 2012
panic(cpu 0 caller 0xffffff8014ab7b95): Kernel trap at 0xffffff7f962e6821, type 13=general protection, registers:
CR0: 0x000000008001003b, CR2: 0x0000000106861000, CR3: 0x0000000020979000, CR4: 0x0000000000000660
RAX: 0xdeadbeefdeadbeef, RBX: 0xffffff8023edb0b8, RCX: 0xffffff802421a228, RDX: 0xffffff802421a368
RSP: 0xffffff808de93a20, RBP: 0xffffff808de93a40, RSI: 0x00000000deadbeef, RDI: 0xffffff8020ffcf00
R8: 0xffffff80150bec60, R9: 0x00007fff5527b2d0, R10: 0x000000000000002c, R11: 0x0000000000000206
R12: 0xffffff807f4ed000, R13: 0x00000000000002aa, R14: 0xffffff8020ffcf00, R15: 0x00000000051ca000
RFL: 0x0000000000010282, RIP: 0xffffff7f962e6821, CS: 0x0000000000000008, SS: 0x0000000000000000
Fault CR2: 0x0000000106861000, Error code: 0x0000000000000000, Fault CPU: 0x0

Backtrace (CPU 0), Frame : Return Address
0xffffff808de936c0 : 0xffffff8014a1d5f6
0xffffff808de93730 : 0xffffff8014ab7b95
0xffffff808de93900 : 0xffffff8014ace4ad
0xffffff808de93920 : 0xffffff7f962e6821
0xffffff808de93a40 : 0xffffff7f96331571
0xffffff808de93a80 : 0xffffff7f962eb2fd
0xffffff808de93aa0 : 0xffffff7f962e4235
0xffffff808de93af0 : 0xffffff7f962e2dba
0xffffff808de93b10 : 0xffffff7f962e4e30
0xffffff808de93b40 : 0xffffff8014e63e8b
0xffffff808de93bc0 : 0xffffff8014e65eb3
0xffffff808de93c20 : 0xffffff8014e638ef
0xffffff808de93d70 : 0xffffff8014a981a1
0xffffff808de93e80 : 0xffffff8014a20abd
0xffffff808de93eb0 : 0xffffff8014a10448
0xffffff808de93f00 : 0xffffff8014a195fb
0xffffff808de93f70 : 0xffffff8014aa5ad6
0xffffff808de93fb0 : 0xffffff8014aced13
Kernel Extensions in backtrace:
com.apple.ATIRadeonX2000(8.0)[F2643601-396F-313A-8D94-5A36091345D0]@0xffffff7f962c7000->0xffffff7f96592fff
dependency: com.apple.iokit.IOPCIFamily(2.7)[8C9E06A4-13D0-33F5-A377-9E36F0ECC229]@0xffffff7f9508a000
dependency: com.apple.iokit.IONDRVSupport(2.3.4)[E37F420A-B5CD-38ED-9441-5247583B6ACE]@0xffffff7f955d1000
dependency: com.apple.iokit.IOGraphicsFamily(2.3.4)[5D671681-D21B-3CCA-9810-B15E648C1B27]@0xffffff7f9558e000

BSD process name corresponding to current thread: WindowServer

kernel panic

I’m sure y’all remember the iMac freezes, I whined about them here a lot. Currently I’m running with unmodified kernel extensions because of two reasons. First, the freezes are infrequent enough to be still bearable, albeit they do come at the worst possible moments, adhering to the laws of the universe. Second, the feature introduced in Lion that resumes all running apps in the place and state they were when the shutdown was issued makes restarts much less of an issue than before.

Today I was reading an article, when… This time I wanted to do things right so I fired up the old white MacBook “Snowflake” and used ssh to log in to the iMac, then dmesg to confirm that it was, in fact, the usual video freeze: for sure the kernel output had a nice long GPU dump in it. I promptly (pun intended) issued a reboot now command only to be greeted by this:

Well, that’s a first. Never before did I get a kernel panic screen.

And Apple claims it is a hardware issue.

I wonder if Mountain Lion will be any better.

iMac screen freezes caused by faulty hardware

The title basically spoils the ending, but here’s how my dealings with Apple went regarding the freezing iMac.

To recap the issue: ever since 10.6.3 some Macs produce video freezes where the computer remains running but cannot be interacted with. Screen can black out, white out, show stripes or just an endless beachball. The only solution is to hard reset or – occasionally – use ssh to log into the machine and issue a reboot command. The kernel.log is usually flooded with a particularly disgusting GPU dump.

There is a fairly extensive discussion about this problem on the Apple Support Communities board which started in March 201024″ iMac Screen Freezes since 10.6.3 update – pls help!

People in the thread determined that something in the video drivers cause the freezes and that the relevant kernel extensions can be swapped to their 10.6.2 counterparts which stops the lockups. Of course one loses two years worth of video driver updates this way. Regardless, I’ve been doing this for every OS X update since then.

Feeling that I exhausted all available options one day in February I was just fed up enough to exasperatedly email Tim Cook. After all some people did get a reply…

To my huge surprise I did actually receive a response from an Executive Relations representative who forwarded the issue to a Senior Apple Care Specialist. Although the warranty on the iMac has long been expired, they agreed to deal with the case based on the assumption that it is a software problem. I was quite hopeful we could finally find out the cause of this issue together.

Over the phone I was instructed to set up a separate partition with a fresh install of OS X Lion to prove that none of my installed third party software are the culprit. For weeks I was trying to reproduce the freeze on that installation with little success almost losing belief that it would occur until finally it did freeze out on me, classic rainbow cursor style. Logs and system information were gathered and sent over in hope.

Only the reply I got a week later was a major letdown: the engineers allegedly determined that the root of the issue is malfunctioning hardware. I was told that since the warranty had expired they are unable to offer a replacement video card or any other free solution. What’s infuriating about this is that my iMac is an early 2008 model; 10.6.3 came out on 29 March 2010, meaning I was most likely out of the 1 year warranty when I even had the chance to find out about the problem. Along with several others I was sold a computer with allegedly faulty hardware, with said fault well hidden during the time covered by warranty. That, or everyone’s video cards just fried the day 10.6.3 came out…

I seem to recall that replacement programs have been started for much less. Also we are talking about a company with enough cash to buy my home country’s national debt. And I still get to use a computer with a broken video card.

10.7.3

Kijött a 10.7.3 Lion update. Én jelenleg 10.7.2-t futtatok 10.6.2-es kext fileokkal, azaz kernel extensionökkel, amik a video megjelenítésért felelnek. A probléma dióhéjban: egyes iMac gépek látszólag abszolút véletlenszerűen képernyő-fagyásokat produkálnak a 10.6.3-as frissítés óta. Ilyenkor a végtelen standlabdázástól a szürke képernyőn át a csíkozásig bármi lehet. A GPU csodás dumpot hány a syslogba, a gépet pedig jó esetben ssh-n keresztül, rossz esetben hardveresen lehet újraindítani.

Most akkor megint megy a tesztelés pár napig, hogy az új frissítéssel jön-e a fagyás. Az Apple pedig magasan tojik a fejünkre, hiszen egyrészt már nem kapható modellekről van szó, másrészt a userbase kis százalékát érinti.

Márpedig ez akkor is egy szoftveres probléma, más nem lehet.

How to brick an iMac

A mai nap tanulsága: az ember ne játsszon a kernel extensionökkel, avagy hogyan tegyünk tönkre egy iMacet.

A hozzám hasonlóan állandó grafikai fagyásokkal megáldott iMac felhasználók titkon azért reménykedtek, hátha belecsempésznek egy javítást a 10.6.6 frissítésbe. Nem így történt: ma délelőtt kétszer egymás után változatos színűre fagyott a kijelző.

Az Apple Support fórumon egy csávó azt állítja, hogy a 10.6.2-es patchből kioperált meghajtókkal neki sikerült megszüntetnie ezt az iszonyatosan bosszantó jelenséget. Csupán három kernel extensiönt kell kicserélni…

A meghajtók lecserélése után az iMac megragadt az alma logós boot forgó kerekénél. Újraindítottam parancssoros módban, és visszacseréltem az ATI2600 kext filet. A support és framebuffer meghajtók a régiek maradtak. Lehet, hogy itt követtem el a hibát. Vagy ott, hogy nem használtam a kext helpert.

Ugyanis ettől a pillanattól fogva az iMac nem bootol, bármit csinálok vele. SMC reset nem segít, NVRAM reset nem segít. NVRAM reset háromszor egymás után nem segít. Safe modeba sem bootol, CD-ről sem bootol. Tritone, és fehér/szürke képernyő. Ennyit tud a gép jelenleg.

Tesóm hazavitte a MacBook töltőmet, a laptopon meg jó ha 5% akkuidő maradt. (A sajátjának elégette a vezetékét. Don’t ask.) Az ubuntu szerveremhez nincs monitorom. Most elloptam kölcsönvettem tesóm monitorját és billentyűzetét, úgy írom ezt a bejegyzést.

Kissé kétségbe vagyok esve. Az iMacet javítani valószínűleg többhavi fizetésem lenne. Ötletem sincs, mit csináljak.

Update: Kipróbáltam a Target Disk Modeot, természetesen nem működik. Az elején azt hittem, abszolút semmit sem csinál, de pár perc állás után megjelent a Firewire ikon. Csak épp a hoszt gép nem lát belőle semmit. Ha kihúzom majd bedugom a vezetéket, akkor pedig ezt írja a konzol:

Jan 17 18:34:30 Snowflake SystemUIServer[112]: ICANotifications framework timed out waiting for a FireWire device with GUID ‘9907699246124904’ to become ready!

A furcsa az iMac viselkedésében az, hogy az amúgy a képernyő szélei között pattogó Firewire ikon (megnéztem, hogy kellene kinéznie a MacBookon) percenként ugrik egy framet. Mintha valami működésképtelenre lassítaná a gépet.

Ugyanez a CD is. Nem bootol az OS X DVD-ről, de vagy öt percig szenved magában, mire kiadja a DVD-t. Erre is véletlenül jöttem rá, amikor egyszer dühömben úgy hagytam.

Egyre gyanúsabb, hogy ez nem szimpla HDD probléma, és keresztet vethetek az egész gépre.

2010

On 31 December 2010 at 22.40 while I was sitting alone at home the iMac produced one of its usual screen freezes.

Seriously, fuck this year.

24" iMac screen freeze

As some of the readers probably know I am a proud owner of a 2008 aluminium iMac (model 8,1).

However, lately I am being plagued by a very annoying and – so far – unavoidable issue: during use at a random point the screen just freezes, something like a permanent screenshot. I can’t tell when it happens, can be weeks without it coming up at all or just after a few hours of being switched on. Sometimes the mouse still works but no input is registered. The machine usually keeps running and – for example if iTunes was on – music keeps playing in the background. I can use ssh to log in to the machine as well.

In order to put the machine out of its misery the only solution is a hard reset. (Or issuing a reboot/shutdown command from ssh.)

I can’t really tell when the phenomenon started, could be around the 10.6.3 update.

I have tried the following methods so far:

So far none of the above helped and I am getting increasingly frustrated.

Some users reported that their video card/logic board was replaced by Apple, only there is no store in Hungary and I’m out of warranty… Regardless, I can’t tell if it is a software issue or a hardware issue brought up by software updates.

Update: So far it seems that Time Machine and the Firewire drive has to do something with the freezes. Unplugged it and freezing ceased, replugged as USB and I got a lockup within the day. Doing a long-term test now with no TM/drive.

Update 2: I have installed 10.6.5 yesterday, replugged the TM drive and only this morning I got another freeze. Apparently the patch did nothing to fix the problem. Testing with unplugged external storage once again.

Update 3: OS X just got endless beachballed while no TM drive was present. Back to square one. Apparently I’m not alone either.