Jump to content

Mavericks kernel testing on AMD (formerly Mountain Lion kernel testing on AMD)

theconnactic

By theconnactic,
in OS X 10.9 (Mavericks)

 Share
6,414 posts in this topic

Recommended Posts

hzhjy123
hzhjy123
Posted
Posted

Thanks！！！ @Andy @Duran Keeley  
My card is working now!! And system is working SO GOOD!
Because I am from China,So my system language is Chinese.
Give some photos!file:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.11.47.pngfile:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.11.47.png

file:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.10.29.png
file:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.10.32.png
file:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.11.35.png

sorry i don't how to upload photos


Sorry,it's here ;) post-1160681-0-10280800-1383440104_thumb.pngpost-1160681-0-14059500-1383440142_thumb.pngpost-1160681-0-25771300-1383440197_thumb.pngpost-1160681-0-98383100-1383440230_thumb.pngpost-1160681-0-79103600-1383440268_thumb.png

Link to comment
Share on other sites
+ira
+ira
Posted
Posted

The most baffling thing I can remember about the nVidia bug was the following behavior: whenever I started up Mountain Lion after previously restarting from i386 Lion, the screen would display a distorted image, or even quite clear pieces of, the Lion desktop and all open windows - like it was simply dumping onscreen the contents of VRAM.

That's so weird I've never faced this kind of symptom. But I do remember one weird thing on some Bronya 10.8 kernel release, the screen was redered with a lot of graphics glitches/artifacts but as soon as the color management profile changed they all disappear. Also in 10.7 the rendering performance is directly affected by the color profile in use...

  • Like 1
Link to comment
Share on other sites
Diazz72
Diazz72
Posted
Posted

Thanks！！！ @Andy @Duran Keeley  

My card is working now!! And system is working SO GOOD!

Because I am from China,So my system language is Chinese.

Give some photos!file:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.11.47.pngfile:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.11.47.png

file:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.10.29.png
file:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.10.32.png
file:///Users/hjy/Desktop/屏幕快照%202013-11-03%20上午8.11.35.png

sorry i don't how to upload photos

Sorry,it's here ;) attachicon.gif屏幕快照 2013-11-03 上午8.10.19.pngattachicon.gif屏幕快照 2013-11-03 上午8.10.29.pngattachicon.gif屏幕快照 2013-11-03 上午8.10.32.pngattachicon.gif屏幕快照 2013-11-03 上午8.11.35.pngattachicon.gif屏幕快照 2013-11-03 上午8.11.47.png

Can you give some info on it ;). . .Tnx

Link to comment
Share on other sites
Chris Etheridge
Chris Etheridge
Posted
Posted

That's so weird I've never faced this kind of symptom. But I do remember one weird thing on some Bronya 10.8 kernel release, the screen was redered with a lot of graphics glitches/artifacts but as soon as the color management profile changed they all disappear. Also in 10.7 the rendering performance is directly affected by the color profile in use...

I've been getting this with Mavericks! It would dump the desktop screen, all distorted, from my main partition being lion, and after about 3 minutes it would clear and load the login screen to Mavericks
Link to comment
Share on other sites
Adrian Andréas
Adrian Andréas
Posted
Posted

The most baffling thing I can remember about the nVidia bug was the following behavior: whenever I started up Mountain Lion after previously restarting from i386 Lion, the screen would display a distorted image, or even quite clear pieces of, the Lion desktop and all open windows - like it was simply dumping onscreen the contents of VRAM.

The same happened to me

however that my card is ATI!

Link to comment
Share on other sites
Andy Vandijck
Andy Vandijck
Posted
Posted

Maybe is this problem and reason for artifacts. . .

No it is part of RadeonSensors (FakeSMC).

I could give you a version that can as I modified mine to wait longer for the IOAccelerator and not print that annoying still waiting for IOAccelerator statement.

  • Like 1
Link to comment
Share on other sites
Andy Vandijck
Andy Vandijck
Posted
Posted

I've currently arrived at my home and am making the mods we discussed yesterday.

I've done the CPU features bits fix and also added to that same bit of code the setting of PENRYN CPU family for any Intel CPU without SSSE3 (64-bit Pentium 4 maybe?)

I'm after that going to make the blockkexts boot argument in order to block user specified extensions...

Keep you posted ;)

Code for features fix:

    /* AnV - Fix cpuid features bit for SSE3 / SSSE3 and CPU family to PENRYN for Intel too (Pentium 4 fix) - Thanks to Sinetek for the suggestion... */
    if (!(info_p->cpuid_features & CPUID_FEATURE_SSE3) || !(info_p->cpuid_features & CPUID_FEATURE_SSSE3))
    {
        info_p->cpuid_features |= CPUID_FEATURE_SSE3;
        info_p->cpuid_features |= CPUID_FEATURE_SSSE3;
        /* Needed for CPU's without SSSE3 like 64-bit Pentium 4 */
        info_p->cpuid_cpufamily = CPUFAMILY_INTEL_PENRYN;
    }
  • Like 2
Link to comment
Share on other sites
granik
granik
Posted
Posted

I've currently arrived at my home and am making the mods we discussed yesterday.

I've done the CPU features bits fix and also added to that same bit of code the setting of PENRYN CPU family for any Intel CPU without SSSE3 (64-bit Pentium 4 maybe?)

I'm after that going to make the blockkexts boot argument in order to block user specified extensions...

Keep you posted  ;)

 

very old AMD too ?
  • Like 1
Link to comment
Share on other sites
awkimballjr
awkimballjr
Posted
Posted

So I haven't posted here before but I've been following this topic and testing the kernels and decided to share my input and join the testing team!

 

My specs are:

AMD FX-8320 at 3.5ghz

MSI 970a-G46 AM3+ motherboard

Sapphire Radeon HD7850 2GB OC GPU

8GB of DDR3 RAM

 

I'm currently running Bronya's RC5 with SECURE test kernel and it's really stable. I get graphics issues in Safari, Chrome, the App Store, and sometimes Finder. Running Cinebench I got a 6.9 in the CPU but for my GPU I got the red pixel render error and it didn't give me a score.

 

Link to comment
Share on other sites
Andy Vandijck
Andy Vandijck
Posted
Posted

 

very old AMD too ?

 

Any AMD gets automatically PENRYN family class...

For CPUs without SSSE3 it sets the required feature bit (as needed by the vector code as pointed out by Sinetek).

I'm currently working on the custom blacklister.

The argument will be blockkexts.

It will support up to 16 custom blocked kexts.

Simple and efficient function to copy the name of 1 function over:

/* AnV - This function propagates a custom blacklist argument */
void SetBlacklistArg(char **out, char *name, uint32_t start, uint32_t stop)
{
    int i = 0;
    uint32_t current = start;

    while ((i < 256) && (current < stop))
    {
        *out[i] = name[current];

        ++i;
        ++current;
    }
}

Function for getting the arguments in a list:

    char                confArgs[4120];
    uint32_t            confStart = 0;
    uint32_t            confStop = 0;
    uint32_t            confLen = 0;
    uint32_t            confCur = 0;

    if (PE_parse_boot_argn("blockkexts", confArgs, sizeof(confArgs)))
    {
        confLen = strlen(confArgs);

        while (confCur < confLen)
        {
            if ((confArgs[confCur] == ',') && (confblacklistCount < 16))
            {
                confStop = confCur - 1;
                SetBlacklistArg((char **)&(confblacklistMods[confblacklistCount].name), confArgs, confStart, confStop);
                ++confblacklistCount;
                confStart = confCur + 1;
            }

            ++confCur;
        }

        if (confLen == 0)
        {
            confblacklistEnabled = FALSE;
        } else {
            if ((confblacklistCount < 16) && (confStart < confLen))

            {
                confStop = confLen;
                SetBlacklistArg((char **)&(confblacklistMods[confblacklistCount].name), confArgs, confStart, confStop);
                ++confblacklistCount;
            }

            confblacklistEnabled = TRUE;
        }
    }

In the header I defined these:

    typedef struct {
        char name[256];
        uint32_t hits;
    } blacklist_confmod_t;
    extern boolean_t confblacklistEnabled;
    extern blacklist_confmod_t confblacklistMods[16];
    extern uint32_t confblacklistCount;

As you can see, the ',' is used as delimiter...

So you can for example pass the following argument:

blockkexts=com.apple.driver.AppleHDAController,com.apple.driver.AppleEFIRuntime

which blocks AppleHDA.kext and AppleEFIRuntime.kext.

Now on to the blocking code... ;)

EDIT: Oh yeah, I built the latest Clover (2284) from source for those who want it. (BiosBlockIO + SATA version + 32bit version)

Attached below

MD5 of PKG: f417f00ae2898c5d024a8c23e1ed89f8

 

EDIT:

Ok, got it done I think

Last bit of code in IOCatalogue.cpp added:

            if ((confblacklistEnabled) && (confblacklistCount > 0)) {
                OSString *modName = OSDynamicCast(OSString, dict->getObject(gIOModuleIdentifierKey));
                const char *modNameStr = NULL;
                if (modName)
                    modNameStr = modName->getCStringNoCopy();
                if (modNameStr) {
                    boolean_t shouldMatch = TRUE;
                    for (uint32_t n = 0; n >= confblacklistCount; n++) {
                        if (strcmp(confblacklistMods[n].name, modNameStr))
                           continue;
                        if (!confblacklistMods[n].hits++)
                            printf("warning: skipping personalities in user blacklisted kext %s\n", modNameStr);
                        shouldMatch = FALSE;
                }
                if (!shouldMatch)
                    continue;
            }

Kernel is building now... :D

Clover_v2k_r2284.pkg.zip

  • Like 4
Link to comment
Share on other sites
tragediana150
tragediana150
Posted
Posted

As you can see, the ',' is used as delimiter...

So you can for example pass the following argument:

blockkexts=com.apple.driver.AppleHDAController,com.apple.driver.AppleEFIRuntime

which blocks AppleHDA.kext and AppleEFIRuntime.kext.

Now on to the blocking code... ;)

Brilliant!! :D:)

Link to comment
Share on other sites
Andy Vandijck
Andy Vandijck
Posted
Posted

Ok, build done...

Going to test it first myself, if it works right, upload ;)

EDIT: But first some food, need to eat too lol :D

Link to comment
Share on other sites
Andy Vandijck
Andy Vandijck
Posted
Posted

Do you need access to a AMD machine for testing? Because I set up remote access for devs.

No going to get one from a friend next week.

 

I'm ironing out the last bugs.

Almost have it working ;)

Link to comment
Share on other sites
 Share
Go to topic listing
×
×
  • Create New...