boot ok,but reboot hang, from 2.6.10 to 2.6.14
when I reboot from linux(not from windows or BIOS,they all reboot ok from them),it will hang after the system POST before enter the grub menu list
This only happens in reboot from linux kernel within 2.6.10-2.6.14.
I tested the older kernel version from 2.6.8 to 2.6.14,and 2.4.31 .I found the 2.6.8 and the 2.6.9,2.4.31 worked well without above reboot_from_linux_with_hang_after_POST,and the 2.6.10-2.6.14 all have this.
Did this belong to the special reboot_ fixed option in the kernel?
I send to the LKML before ,but it seemed like none fixed it
Quoted 7 lines · Boxer Gnome
This only happens in reboot from linux kernel within 2.6.10-2.6.14.
I tested the older kernel version from 2.6.8 to 2.6.14,and 2.4.31 .I found the 2.6.8 and the 2.6.9,2.4.31 worked well without above reboot_from_linux_with_hang_after_POST,and the 2.6.10-2.6.14 all have this.
Can you try to pinpoint when it started happening more closely?
The differences between 2.6.9 and 2.6.10 are pretty big, and it would be much better if you can pinpoint it to a smaller range.
You can find three "release candidates" for 2.6.10 in
http://www.kernel.org/pub/linux/kernel/v2.6/testing
and if you first test 2.6.10-rc2, and then depending on whether that already has the bug or not, you'd test 2.6.10-rc1 or 2.6.10-rc2. That would help pinpoint the difference to between two particular -rc kernels, which would be much better.
After that, I might end up still asking you to test one or two daily snapshots, but it may be that pinpointing when your reboot troubles started to just the -rc kernel might be good enough.
Linus
2005/10/28, Linus Torvalds <[email protected]>:
Quoted 25 lines · Boxer Gnome
Quoted 7 lines
This only happens in reboot from linux kernel within 2.6.10-2.6.14.
I tested the older kernel version from 2.6.8 to 2.6.14,and 2.4.31 .I found the 2.6.8 and the 2.6.9,2.4.31 worked well without above reboot_from_linux_with_hang_after_POST,and the 2.6.10-2.6.14 all have this.
Can you try to pinpoint when it started happening more closely?
The differences between 2.6.9 and 2.6.10 are pretty big, and it would be much better if you can pinpoint it to a smaller range.
You can find three "release candidates" for 2.6.10 in
http://www.kernel.org/pub/linux/kernel/v2.6/testingand if you first test 2.6.10-rc2, and then depending on whether that already has the bug or not, you'd test 2.6.10-rc1 or 2.6.10-rc2. That would help pinpoint the difference to between two particular -rc kernels, which would be much better.
Quoted 6 lines
After that, I might end up still asking you to test one or two daily snapshots, but it may be that pinpointing when your reboot troubles started to just the -rc kernel might be good enough.
Linus
I tested the 2.6.10-rc1 and 2.6.10-rc2,the 2.6.10-rc1 rebooted ok,but the 2.6.10-rc2 has that problem.
Then I test the snapshot 2.6.10-rc2-bk1,2.6.10-rc2-bk4,2.6.10-rc2-bk8,they all have this reboot hang.
SO,I think the 2.6.10-rc1 is the last worked ok version.
Thanks
Quoted 3 lines · Boxer Gnome
I tested the 2.6.10-rc1 and 2.6.10-rc2,the 2.6.10-rc1 rebooted ok,but the 2.6.10-rc2 has that problem.
Thanks. So it's between 2.6.10-rc1 and -rc2. Good. However:
Quoted 3 lines
Then I test the snapshot 2.6.10-rc2-bk1,2.6.10-rc2-bk4,2.6.10-rc2-bk8,they all have this reboot hang.
Actually, you should test 2.6.10-rc1-bk*, not -rc2-bk*.
The "-bkxxx" kernels are not release-candidates, so they are not "before" the real release. They are nightly snapshots _after_ the release, so 2.6.10-rc2-bk1 is the first snapshot after 2.6.10-rc2.
So you'd be interested in the ones after -rc1, but before -rc2.
Linus
2005/10/29, Linus Torvalds <[email protected]>:
Quoted 15 lines · Boxer Gnome
Quoted 3 lines
I tested the 2.6.10-rc1 and 2.6.10-rc2,the 2.6.10-rc1 rebooted ok,but the 2.6.10-rc2 has that problem.
Thanks. So it's between 2.6.10-rc1 and -rc2. Good. However:
Quoted 3 lines
Then I test the snapshot 2.6.10-rc2-bk1,2.6.10-rc2-bk4,2.6.10-rc2-bk8,they all have this reboot hang.
Actually, you should test 2.6.10-rc1-bk*, not -rc2-bk*.
Quoted 8 lines
The "-bkxxx" kernels are not release-candidates, so they are not "before" the real release. They are nightly snapshots _after_ the release, so 2.6.10-rc2-bk1 is the first snapshot after 2.6.10-rc2.
So you'd be interested in the ones after -rc1, but before -rc2.
Linus
OK,I'd tested the 2.6.10-rc1-bk1,2.6.10-rc1-bk6,2.6.10-rc1-bk12,I found the 2.6.10-rc1-bk1, 2.6.10-rc1-bk6 rebooted ok,but the 2.6.10-rc1-bk12 has that problem.
Then I tested the 2.6.10-rc1-bk8,2.6.10-rc1-bk9,2.6.10-rc1-bk10.
The 2.6.10-rc1-rc8 was still ok.the 2.6.10-rc1-bk9 began has the reboot_hang_after_POST.
Hope this help you.
Thanks
Sorry,I typed a little error.
the"The 2.6.10-rc1-rc8 was still ok" should is "The 2.6.10-rc1-bk8 was still ok"
2005/10/29, Boxer Gnome <[email protected]>:
Quoted 38 lines
2005/10/29, Linus Torvalds <[email protected]>:
Quoted 15 lines · Boxer Gnome
Quoted 3 lines
I tested the 2.6.10-rc1 and 2.6.10-rc2,the 2.6.10-rc1 rebooted ok,but the 2.6.10-rc2 has that problem.
Thanks. So it's between 2.6.10-rc1 and -rc2. Good. However:
Quoted 3 lines
Then I test the snapshot 2.6.10-rc2-bk1,2.6.10-rc2-bk4,2.6.10-rc2-bk8,they all have this reboot hang.
Actually, you should test 2.6.10-rc1-bk*, not -rc2-bk*.
Quoted 8 lines
The "-bkxxx" kernels are not release-candidates, so they are not "before" the real release. They are nightly snapshots _after_ the release, so 2.6.10-rc2-bk1 is the first snapshot after 2.6.10-rc2.
So you'd be interested in the ones after -rc1, but before -rc2.
Linus
OK,I'd tested the 2.6.10-rc1-bk1,2.6.10-rc1-bk6,2.6.10-rc1-bk12,I found the 2.6.10-rc1-bk1, 2.6.10-rc1-bk6 rebooted ok,but the 2.6.10-rc1-bk12 has that problem.
Then I tested the 2.6.10-rc1-bk8,2.6.10-rc1-bk9,2.6.10-rc1-bk10.
The 2.6.10-rc1-rc8 was still ok.the 2.6.10-rc1-bk9 began has the
reboot_hang_after_POST.Hope this help you.
Thanks
Quoted 4 lines · Boxer Gnome
Sorry,I typed a little error.
the"The 2.6.10-rc1-rc8 was still ok" should is "The 2.6.10-rc1-bk8 was still ok"
Ok, thanks, you've been a hero.
This cuts the changes down quite significantly. The shortlog between rc1-bk8 and rc1-bk9 is (if I did everything right - since that's from the BK days and I obviously don't have BK installed anywhere, I had to convert things into git to make sense of it) appended.
I _suspect_ it's the IDE changes: perhaps the fact that we enable IDE "stroke" by default now, which unclips the IDE disk. Maybe the BIOS is unhappy with that.
In your "dmesg" for a modern kernel, do you get a message like
...: Host Protected Area detected.
current capacity is .. sectors (.. MB) native capacity is .. sectors (.. MB)
if so, that's probably it, and you could try this patch.
If that doesn't help, I can help you bisect the problem down to a smaller patch, if you're willing to try some patches I'd send by email?
Linus
---
Hacky patch for you to try with current kernel:
drivers/ide/ide-disk.c +2 -0
@@ -498,6 +498,8 @@ static inline void idedisk_check_hpa(ide capacity,sectors_to_MB (capacity), set_max, sectors_to_MB(set_max)); +return ; +if (lba48) set_max =idedisk_set_max_address_ext (drive, set_max);else
---
shortlog between rc1-bk8 and rc1-bk9:
Andi Kleen:
x86_64: Fix safe_smp_processor_id after genapic
x86_64: Fix warning in genapic
Andrew Morton:
key_init ordering fix
swapper_space warning suppression
ext3 reservation: default to on
convert pipefs to fs_initcall()
Antonino Daplas:
fbdev: Convert MODULE_PARM to module_param in i810fb
fbdev: Remove module parameter 'disabled' from savagefb
fbdev: Convert MODULE_PARM to module_param in intelfb
fbdev: Convert MODULE_PARM to module_param in neofb
fbdev: Fix io access in neofb
fbdev: Add __iomem annotations to sstfb
fbdev: Add __iomem annotations to tdfxfb
fbdev: Do not memset the framebuffer memory in asiliantfb
fbdev: Add __iomem annotations to cyber2000fb
fbdev: Add __iomem annotations to pm2fb
fbdev: Add __iomem annotations to hgafb
fbdev: Add __iomem annotations to cirrusfb
fbdev: Add __iomem annotations to vfb
fbdev: Check if cursor image has changed in intelfb
fbdev: Maintainership
Bartlomiej Zolnierkiewicz:
[ide] PIO bugfix
[ide] remove hwif from /proc/ide/ as part of ide_unregister_hwif() [ide] hpt34x: kill hpt34x.h
[ide] pmac: kill CONFIG_BLK_DEV_IDEDMA_PMAC_AUTO
[ide] setup-pci: small ide_get_or_set_dma_base() cleanup [ide] setup-pci: simplify autodma logic
[ide] kill IDEPCI_FLAG_FORCE_MASTER
[ide] make destroy_proc_ide_interfaces static
[ide] ide-disk: enable stroke by default
Benjamin Herrenschmidt:
ppc32: Fix boot on PowerMac
ppc64: Enable maple IDE fixup
Chris Wright:
uninline __sigqueue_alloc
Daniel Jacobowitz:
Unwind information fix for the vsyscall DSO
Dmitry Torokhov:
ik8.c: export power_status parameter through sysfs
Hirokazu Takata:
m32r: fix a typo of delay.c
James Nelson:
Documentation/cpqarray.txt update
Documentation/mkdev.ida removal
Linus Torvalds:
Make x86 semaphore routines use register calling convention.
Merge bk://bart.bkbits.net/ide-2.6
Merge bk://bk.arm.linux.org.uk/linux-2.6-rmkMarcel Holtmann:
Fix deprecated MODULE_PARM for CAPI subsystem
Matt Porter:
ppc32: fix ppc4xx_progress warnings
Mingming Cao:
ext3 block reservation patch set -- ext3 preallocation cleanup ext3 block reservations
Olof Johansson:
ppc64: setup cpu_sibling_map on iSeries
Paolo \'Blaisorblade\' Giarrusso:
uml: fix mainline lazyness about TTY layer patch
Pavel Machek:
Add typechecking to suspend types and powerdown types
Robert Love:
make dnotify a configure-time option
Russell King:
[ARM] Use cpu_vm_mask to indicate whether the MM is mapped. [ARM] Use cpu_vm_mask to determine whether to flush TLB/caches. [ARM] Add disable_irq_nosync() and CPU number headings [ARM] Remove extraneous spaces.
[ARM] include/asm-arm/arch-integrator/time.h is unused, remove it. [ARM] Fix wrong variable name in icside.c
Stephen C. Tweedie:
ext3: online resizing
Stephen Rothwell:
ppc64 iSeries: fix for generic irq changes
Toshihiro Iwamoto:
direct IO write memory leak fix