2.6.18-rc4-no2 "To be..." | (Super Fix/Stability Attempt?)

[zxy]

Cannot apply patch.

Code: Select all

cat patches/time-patch | patch -p1
patching file kernel/time/ntp.c
Hunk #1 FAILED at 41.
1 out of 1 hunk FAILED -- saving rejects to file kernel/time/ntp.c.rej

here is kernel/time/ntp.c.rej

Code: Select all

***************
*** 41,47 ****
  long time_adjust;

  #define CLOCK_TICK_OVERFLOW   (LATCH * HZ - CLOCK_TICK_RATE)
- #define CLOCK_TICK_ADJUST   (((s64)CLOCK_TICK_OVERFLOW * NSEC_PER_SEC) / CLOCK_TICK_RATE)

  static void ntp_update_frequency(void)
  {
--- 41,47 ----
  long time_adjust;

  #define CLOCK_TICK_OVERFLOW   (LATCH * HZ - CLOCK_TICK_RATE)
+ #define CLOCK_TICK_ADJUST   (((s64)CLOCK_TICK_OVERFLOW * NSEC_PER_SEC) / (s64)CLOCK_TICK_RATE)

  static void ntp_update_frequency(void)
  {

I changed the line myself, added (s64).
I do not have the

Code: Select all

 long time_adjust;

there
Here is what is there in the file

Code: Select all

/**
 * ntp_clear - Clears the NTP state variables
 *
 * Must be called while holding a write on the xtime_lock
 */
void ntp_clear(void)
{
    time_adjust = 0;        /* stop active adjtime() */
    time_status |= STA_UNSYNC;
    time_maxerror = NTP_PHASE_LIMIT;
    time_esterror = NTP_PHASE_LIMIT;

    ntp_update_frequency();

    tick_length = tick_length_base;
    time_offset = 0;
}

#define CLOCK_TICK_OVERFLOW (LATCH * HZ - CLOCK_TICK_RATE)
#define CLOCK_TICK_ADJUST   (((s64)CLOCK_TICK_OVERFLOW * NSEC_PER_SEC) / (s64)CLOCK_TICK_RATE)

void ntp_update_frequency(void)
{
    tick_length_base = (u64)(tick_usec * NSEC_PER_USEC * USER_HZ) << TICK_LENGTH_SHIFT;
    tick_length_base += (s64)CLOCK_TICK_ADJUST << TICK_LENGTH_SHIFT;
    tick_length_base += (s64)time_freq << (TICK_LENGTH_SHIFT - SHIFT_NSEC);

    do_div(tick_length_base, HZ);

    tick_nsec = tick_length_base >> TICK_LENGTH_SHIFT;
}

/*

[cheater1034]

Cannot apply patch.

Code: Select all

cat patches/time-patch | patch -p1
patching file kernel/time/ntp.c
Hunk #1 FAILED at 41.
1 out of 1 hunk FAILED -- saving rejects to file kernel/time/ntp.c.rej

here is kernel/time/ntp.c.rej

Code: Select all

***************
*** 41,47 ****
  long time_adjust;

  #define CLOCK_TICK_OVERFLOW   (LATCH * HZ - CLOCK_TICK_RATE)
- #define CLOCK_TICK_ADJUST   (((s64)CLOCK_TICK_OVERFLOW * NSEC_PER_SEC) / CLOCK_TICK_RATE)

  static void ntp_update_frequency(void)
  {
--- 41,47 ----
  long time_adjust;

  #define CLOCK_TICK_OVERFLOW   (LATCH * HZ - CLOCK_TICK_RATE)
+ #define CLOCK_TICK_ADJUST   (((s64)CLOCK_TICK_OVERFLOW * NSEC_PER_SEC) / (s64)CLOCK_TICK_RATE)

  static void ntp_update_frequency(void)
  {

You may need to apply it manually, nano kernel/time/ntp.c

just find the line, delete it, and replace it with the new

[vipernicus]

Code: Select all

I changed the line myself, added (s64).

Good job! It's good to see people who aren't afraid to get their hands dirty.

[zxy]

Still doesn't work.
I even added (s64) to all three constants in the line.
And added long time_adjust; because it wasnt there.

when I do

Code: Select all

watch date

time pases in 2 second steps. This means every second it grows for 2 seconds.
When I recompiled the kernel the last time now, im starting to get clock skew warnings about time beeing in the future.. I guess the time resets from bios at startup.

Other kernels work ok for me.

I don't really have a clue what to do.

The kernel/time/ntp.c looks like this now

Code: Select all

    tick_length = tick_length_base;
    time_offset = 0;
}

long time_adjust;

#define CLOCK_TICK_OVERFLOW (LATCH * HZ - CLOCK_TICK_RATE)
#define CLOCK_TICK_ADJUST   (((s64)CLOCK_TICK_OVERFLOW * (s64)NSEC_PER_SEC) / (s64)CLOCK_TICK_RATE)

void ntp_update_frequency(void)
{
    tick_length_base = (u64)(tick_usec * NSEC_PER_USEC * USER_HZ) << TICK_LENGTH_SHIFT;
    tick_length_base += (s64)CLOCK_TICK_ADJUST << TICK_LENGTH_SHIFT;
    tick_length_base += (s64)time_freq << (TICK_LENGTH_SHIFT - SHIFT_NSEC);

[a7thson]

I can't help you on ieee1394 without any dmesg or system info

I posted emerge --info, lspci as well as kernel messages, and will reboot into this kernel again tonight and give you dmesg. is there anything else you need?
I did find something though -- no guarantees here, but mm2 has a number of ieee1394 fixes. mm2 is already HUGE

Code: Select all

+the-scheduled-removal-of-drivers-ieee1394-sbp2cforce_inquiry_hack.patch
+ieee1394-sbp2-handle-sbp2util_node_write_no_wait-failed.patch
+ieee1394-sbp2-safer-agent-reset-in-error-handlers.patch
+ieee1394-sbp2-recheck-node-generation-in-sbp2_update.patch
+ieee1394-sbp2-better-handling-of-transport-errors.patch
+ieee1394-sbp2-select-scsi-in-kconfig.patch
+ieee1394-sbp2-update-includes.patch
+ieee1394-sbp2-prevent-rare-deadlock-in-shutdown.patch
+initialize-ieee1394-early-when-built-in.patch

either of these:
ieee1394-sbp2-handle-sbp2util_node_write_no_wait-failed.patch
ieee1394-sbp2-recheck-node-generation-in-sbp2_update.patch
look like best candidates to be related, though I haven't had a chance to browse the source nor to try applying them. I didn't see anything obvious that related to libata being broken with CDRW/DVD at first glance unfortunately. I may try patching rc4-no2 with all the ieee1394 fixes present in -mm2 (tonight) and will see if there is any change.

[cheater1034]

Still doesn't work.
I even added (s64) to all three constants in the line.
And added long time_adjust; because it wasnt there.

when I do

Code: Select all

watch date

time pases in 2 second steps. This means every second it grows for 2 seconds.
When I recompiled the kernel the last time now, im starting to get clock skew warnings about time beeing in the future.. I guess the time resets from bios at startup.

Other kernels work ok for me.

I don't really have a clue what to do.

The kernel/time/ntp.c looks like this now

Code: Select all

    tick_length = tick_length_base;
    time_offset = 0;
}

long time_adjust;

#define CLOCK_TICK_OVERFLOW (LATCH * HZ - CLOCK_TICK_RATE)
#define CLOCK_TICK_ADJUST   (((s64)CLOCK_TICK_OVERFLOW * (s64)NSEC_PER_SEC) / (s64)CLOCK_TICK_RATE)

void ntp_update_frequency(void)
{
    tick_length_base = (u64)(tick_usec * NSEC_PER_USEC * USER_HZ) << TICK_LENGTH_SHIFT;
    tick_length_base += (s64)CLOCK_TICK_ADJUST << TICK_LENGTH_SHIFT;
    tick_length_base += (s64)time_freq << (TICK_LENGTH_SHIFT - SHIFT_NSEC);

Don't add all this out of context, get the original ntp file

apply this, http://kernel.org/pub/linux/kernel/peop ... nups.patch

and then try applying that fix again.

[Dinini]

Thinkpad T22, snd_cs46xx kernel module.

First the good news, suspend and hibernate are working now.
And now the bad news, sound has died. Alsa mixer claims no attached devices. module is loaded.

Kernel 2.6.18-rc4 works. Kernel 2.6.18-rc4-no1 has sound but suspend/hibernate are not working. Kernel 2.6.18-rc4-mm2 (with ntp hotfix patch) produces the same lack of sound. (I'm too lazy to go back and try mm1 again.) I presume one of the added mm patches is what has killed my sound between no1 and no2. Oh well back to vanilla.

Thanks so much for the work on this patchset.

[edited to add link to the hotfix patch for ntp]

For clarity, I patched manually, I first unpacked linux-2.6.17.tar.bz2 then applied patch patch-2.6.18-rc4.bz2 then 2.6.18-rc4-mm1.bz2 followed by 2.6.18-rc4-no2.bz2. If I messed up in the patches please let me know.

[cheater1034]

Thinkpad T22, snd_cs46xx kernel module.

First the good news, suspend and hibernate are working now.
And now the bad news, sound has died. Alsa mixer claims no attached devices. module is loaded.

Kernel 2.6.18-rc4 works. Kernel 2.6.18-rc4-no1 has sound but suspend/hibernate are not working. Kernel 2.6.18-rc4-mm2 (with ntp hotfix patch) produces the same lack of sound. (I'm too lazy to go back and try mm1 again.) I presume one of the added mm patches is what has killed my sound between no1 and no2. Oh well back to vanilla.

Thanks so much for the work on this patchset.

[edited to add link to the hotfix patch for ntp]

If you're using the alsa drivers in the kernel, and the latest! alsa-lib and alsa-headers you should be OK with sound, if you're not using the latest possible lib and headers, or you're using alsa-driver in portage it will not work.

[Dinini]

Alsa is from kernel. I'll see about a new version of the utilities but I have old api set so I'm still confused.

[gringo]

Here it is for 2.6.18-rc4-vanilla:
http://tienstra4.flatnet.tudelft.nl/~ge ... .patch.txt

thanks dude, exactly what i was looking for !

cheers

[zxy]

1. Time problem :
No-go.
Applied patch as you wrote. Still the same.

2. Sky2 driver for marvel network card still works weird, slows network down after some time. (opera freezes, FF needs a year to reload forums).

[chittoor]

Hey i am having problems with my no-sources. I posted it in another thread.
Here is the issue

Kernel: linux-2.6.17-no5
I get the following message after init starts and the system hangs and nothing works
The first few lines are:


BUG: unable to handle kernel NULL pointer dereference at virtual address 00000018
Printing eip:
*pde = 00000000
Oops: 0002 [#1]
8K_STACKS SMP
last sysfs file: /block/hda/hda8/dev


after that there are few call traces. I'm too lazy to write every thing down. Is there a way to copy that info ?
Please let me know if I have post my kernel .config

[cheater1034]

Hey i am having problems with my no-sources. I posted it in another thread.
Here is the issue

Kernel: linux-2.6.17-no5
I get the following message after init starts and the system hangs and nothing works
The first few lines are:


BUG: unable to handle kernel NULL pointer dereference at virtual address 00000018
Printing eip:
*pde = 00000000
Oops: 0002 [#1]
8K_STACKS SMP
last sysfs file: /block/hda/hda8/dev


after that there are few call traces. I'm too lazy to write every thing down. Is there a way to copy that info ?
Please let me know if I have post my kernel .config

hmm, odd, can I see a full dmesg?

[cheater1034]

wait a minute,

you're using an outdated version!

Try 2.6.18-rc4-no2 and report back. 2.6.17-no* was buggy

[chittoor]

Okay. This was a lot pain writing down all hex s and tying again. The top of messg scrolled too fast but it seems ok.

This part is continuation of my last post

Code: Select all

last sysfs file:
Modules linked in:
CPU:        1
EIP:         0060:[<c037069a7>] Not tainted VL1
EFLAGS: 00010097  (2.16.17-no5 #7)
eax: 00000000  ebx: ffffffd8   ecx: 00000000  edx: 02bbecc1
esi: 00000000  edi: ffffffd8  ebp: c14cd90  esp: dfcaffc4
ds: 007b  es: 007b  ss: 0068

Process kacpid (pid:15, ti=dfcae000 task=c14cda90 task.ti=dfcae000)
Stack: c0111ce5 c14cdba0 c140df48 07bbbbd3 00000000 00000000 00000000 c14cda90
          00000071 dfcaf184 c011ef27 00010000 00000264 c14ccc54 c14ccc40 c14ccc4c 00000000 c0123fda 00000001 00000000
          00000000 00010000 00000000 00000000

Call Trace:
<c0111ce5>try_to_work_up+0x2ff/0x300 <c011ef27>do_sigaction+0x116/0x154
<c0123fda>worker_thread+0xb2/0x10c <c0111eaf>default_wake_function+0x0/0x8
<c0123f28> worker_thread+0x0/0x10c <c012635b>kthread+0xc2/0xed
<0126199>kthread+0x0/0xed <c01007ad>kernel_thread_heaper+0x5/0xb

code: 10 29 c6 19 d7 01 74 24 14 11 7c 24 18 8b 44 24 14 8b 54 24 18 89 45 5c 89
         55 60 39 dd 0f 81 02 00 00 8b 54 24 0c 8b 4c 24 10 <89> 53 38 89 4b 3c 8b 4c
        24 08 83 41 18 01 83 51 1c 00 89 59 2c 

EIP: [<c0370699>] schedule+0x375/0x62c SC ESP 00 68 df ca ff 4c
-->irq 1, desc: c012f561,depth: ,count:0, unhandled: 0
->handle_irq(): c012f561, handle_hal_irq+0x0/0x1a7
->chip(): c045c000,0xc045c00
->action():00000000

IRQ_DISABLED Set
unexpected IRQ ttrpat vector 01

I wrote this by hand, errors are possible
Damn after all this
K im trying the new ebuild. will report back. My net goes down in 30 mins, if not able to report in 30 mins. I'll report after 7 hrs

K Installed. All fine till now. Thakq for help

[genestyler]

did LVM2 work ?

[zxy]

Waiting for a new release and waiting fast...

[Non_E]

Hi cheater1034. I have been using 2.6.18-rc4-no2 for about a half day now and it seems to be rock stable. Ati-drivers, truecrypt and xfs work as expected. Do you expect swsusp2 to appear in no-sources in the near future? Thank you for no-sources.
Edit: Typo

[cheater1034]

Hi cheater1034. I have been using 2.6.18-rc4-no2 for about a half day now and it seems to be rock stable. Ati-drivers, truecrypt and xfs work as expected. Do you expect swsusp2 to appear in no-sources in the near future? Than you for no-sources.

swsusp2 is probable in the future, not not in the .18-rc series atleast, -mm currently has tons of extra suspend patches that make it harder to port swsusp2 over.

I've been talking to Nigel about it, seems like whenever he gets time he'll make a port, and hopefully maintain it.

[matobsk]

ndiswrapper-1.23 compiles and works with no-sources-2.6.18-rc4-no2.

No CHECKSUM_HW -> CHECKSUM_PARTIAL errors at all, and no segfaults

[_dA_CyANIDe]

hello,

i got this while linking :

LD init/built-in.o
LD .tmp_vmlinux1
drivers/built-in.o: In function `ondemand_powersave_bias_init':
cpufreq_ondemand.c:(.text.ondemand_powersave_bias_init+0x2d): undefined reference to `cpufreq_frequency_get_table'
drivers/built-in.o: In function `powersave_bias_target':
(.text.powersave_bias_target+0x4e): undefined reference to `cpufreq_frequency_table_target'
drivers/built-in.o: In function `powersave_bias_target':
(.text.powersave_bias_target+0x98): undefined reference to `cpufreq_frequency_table_target'
drivers/built-in.o: In function `powersave_bias_target':
(.text.powersave_bias_target+0xc8): undefined reference to `cpufreq_frequency_table_target'
make: *** [.tmp_vmlinux1] Error 1


Any suggestion?

[seren]

Still doesn't work.
I even added (s64) to all three constants in the line.
And added long time_adjust; because it wasnt there.

when I do

Code: Select all

watch date

time pases in 2 second steps. This means every second it grows for 2 seconds.
When I recompiled the kernel the last time now, im starting to get clock skew warnings about time beeing in the future.. I guess the time resets from bios at startup.

Other kernels work ok for me.

I don't really have a clue what to do.

The kernel/time/ntp.c looks like this now

Code: Select all

    tick_length = tick_length_base;
    time_offset = 0;
}

long time_adjust;

#define CLOCK_TICK_OVERFLOW (LATCH * HZ - CLOCK_TICK_RATE)
#define CLOCK_TICK_ADJUST   (((s64)CLOCK_TICK_OVERFLOW * (s64)NSEC_PER_SEC) / (s64)CLOCK_TICK_RATE)

void ntp_update_frequency(void)
{
    tick_length_base = (u64)(tick_usec * NSEC_PER_USEC * USER_HZ) << TICK_LENGTH_SHIFT;
    tick_length_base += (s64)CLOCK_TICK_ADJUST << TICK_LENGTH_SHIFT;
    tick_length_base += (s64)time_freq << (TICK_LENGTH_SHIFT - SHIFT_NSEC);

im confused.. by default watch uses a 2 second update interval, try

Code: Select all

watch --interval 1 date

[Phenax]

My clock speedhax. Same results in all 2.6.18 patchsets. The only reference I could fine is http://lkml.org/lkml/2006/8/20/19 ((And alongg with my clock everything else is faster than normal, mayeb userspace apps use the clockk. ))

> Also, it stil has the funny fast moving clock on x86-64

[zxy]

@seren
Well, don't be confused. It makes 2s intervals. In my case these intervals pass by in less than a second.
Anyway the time passes by with more than double speed. Hm, maybe I just stumbled upon a WARP DRIVE and I didn't even know it. (Yeeeehaaa)


--------------
Gentoo makes my computer fly ...(even throught time .... )

[zxy]

@Phenax

For me this is the only 2.6.18 kernel with this problem. I tried previous no-sources kernels and they worked ok (all from 2.6.18 series) and viper too. I do not know what makes the clock go crazy with this kernel. (on amd64)