-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of
[email protected]
>Ok due to popular demands here is the slightly fixed patch that works
>on both i386 and x86_64. For the i386 version you must not have
>HIGHMEM64G enabled.
>I just rolled it all into one patch as I'm to lazy to transmit all
>3 of them.
I got
Firmware type: LinuxBIOS
Linux version 2.6.19-smp-gc9976797-dirty (root@lbsrv) (gcc version
4.0.2) #196 6
Command line: earlyprintk=ttyS0,115200 apic=debug pci=noacpi,routeirq
snd-hda-i
BIOS-provided physical RAM map:
BIOS-e820: 0000000000000000 - 0000000000001000 (reserved)
BIOS-e820: 0000000000001000 - 00000000000a0000 (usable)
BIOS-e820: 00000000000c0000 - 00000000000f0000 (usable)
BIOS-e820: 00000000000f0000 - 0000000000100000 (reserved)
BIOS-e820: 0000000000100000 - 00000000c0000000 (usable)
BIOS-e820: 0000000100000000 - 0000000240000000 (usable)
dbgp_num: 0
Found EHCI debug port
bar: 10 offset: 098
bar: 10 offset: 098
dbgp pre-set_fixmap_nocache
PANIC: early exception rip ffffffff809c24b4 error 0 cr2 ffff810000203ff8
Call Trace:
[<ffffffff809c24b4>] __set_fixmap+0x84/0x202
[<ffffffff809c05bb>] early_dbgp_init+0x259/0x55c
[<ffffffff8022d958>] __call_console_drivers+0x64/0x72
[<ffffffff809b242b>] do_early_param+0x0/0x57
[<ffffffff809c0a20>] setup_early_printk+0x162/0x17e
[<ffffffff809b2459>] do_early_param+0x2e/0x57
[<ffffffff8023d051>] parse_args+0x159/0x1f3
[<ffffffff809b24c2>] parse_early_param+0x40/0x4c
[<ffffffff809b88ca>] setup_arch+0x1c1/0x636
[<ffffffff809b2534>] start_kernel+0x55/0x208
[<ffffffff809b2173>] _sinittext+0x173/0x177
RIP __set_fixmap+0x84/0x202
With Greg's USB Debug, host and target can talk.
target with console=ttyUSB0,115200n8
host with cat /dev/ttyUSB0
But if use minicom in host, it will not show '\r', I guess the usb debug
cable eat return char. Greg, Can you add that back in usb_debug by
replacing '\n' with '\r', '\n'?
With Eric code in LinuxBIOS, it will report "No device found in debug
port"
YH
"Yinghai Lu" <[email protected]> writes:
> On 12/7/06, Eric W. Biederman <[email protected]> wrote:
>> Ugh. I'd check the code. But it looks like my tweak to the
>> early fixmap code. But my hunch is that my tweak to __fixmap
>> so that it's pud and pmd were prepopulated didn't take on
>> your build.
>
> I missed some options?
Your or I missed a bug fix/enhancement in there somewhere.
Basically my very early setup of the fixmap failed.
Now. I thought I had that covered by preallocated the pud and the pmd
entries. So the only thing missing was the pte entries.
If that is not a big enough hint I will look into it in a bit...
I'm starting to become a big fan of constant initializers. So our
core subsystems don't need initialization code to be useful. All of
these early things are just a pain.
Eric
On Thu, Dec 07, 2006 at 07:48:17PM -0800, Lu, Yinghai wrote:
> With Greg's USB Debug, host and target can talk.
> target with console=ttyUSB0,115200n8
Ugh, no, never use the usb-serial driver as a console device.
That was a bad hack done as a bet many years ago. For many obvious
reasons it does not work well.
> host with cat /dev/ttyUSB0
> But if use minicom in host, it will not show '\r', I guess the usb debug
> cable eat return char. Greg, Can you add that back in usb_debug by
> replacing '\n' with '\r', '\n'?
The usb-serial console code should handle this, I thought we fixed it a
while ago.
But this kind of interface is not what these devices are good for. They
are for the debug port information, not as a usb-serial console device.
Otherwise they are way too expensive of a device...
thanks,
greg k-h
"Lu, Yinghai" <[email protected]> writes:
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of
> [email protected]
>
>
>>Ok due to popular demands here is the slightly fixed patch that works
>>on both i386 and x86_64. For the i386 version you must not have
>>HIGHMEM64G enabled.
>
>>I just rolled it all into one patch as I'm to lazy to transmit all
>>3 of them.
>
>
> I got
>
> Firmware type: LinuxBIOS
> Linux version 2.6.19-smp-gc9976797-dirty (root@lbsrv) (gcc version
> 4.0.2) #196 6
> Command line: earlyprintk=ttyS0,115200 apic=debug pci=noacpi,routeirq
> snd-hda-i
> BIOS-provided physical RAM map:
> BIOS-e820: 0000000000000000 - 0000000000001000 (reserved)
> BIOS-e820: 0000000000001000 - 00000000000a0000 (usable)
> BIOS-e820: 00000000000c0000 - 00000000000f0000 (usable)
> BIOS-e820: 00000000000f0000 - 0000000000100000 (reserved)
> BIOS-e820: 0000000000100000 - 00000000c0000000 (usable)
> BIOS-e820: 0000000100000000 - 0000000240000000 (usable)
> dbgp_num: 0
> Found EHCI debug port
> bar: 10 offset: 098
> bar: 10 offset: 098
> dbgp pre-set_fixmap_nocache
> PANIC: early exception rip ffffffff809c24b4 error 0 cr2 ffff810000203ff8
>
> Call Trace:
> [<ffffffff809c24b4>] __set_fixmap+0x84/0x202
> [<ffffffff809c05bb>] early_dbgp_init+0x259/0x55c
> [<ffffffff8022d958>] __call_console_drivers+0x64/0x72
> [<ffffffff809b242b>] do_early_param+0x0/0x57
> [<ffffffff809c0a20>] setup_early_printk+0x162/0x17e
> [<ffffffff809b2459>] do_early_param+0x2e/0x57
> [<ffffffff8023d051>] parse_args+0x159/0x1f3
> [<ffffffff809b24c2>] parse_early_param+0x40/0x4c
> [<ffffffff809b88ca>] setup_arch+0x1c1/0x636
> [<ffffffff809b2534>] start_kernel+0x55/0x208
> [<ffffffff809b2173>] _sinittext+0x173/0x177
>
> RIP __set_fixmap+0x84/0x202
Ugh. I'd check the code. But it looks like my tweak to the
early fixmap code. But my hunch is that my tweak to __fixmap
so that it's pud and pmd were prepopulated didn't take on
your build.
> With Eric code in LinuxBIOS, it will report "No device found in debug
> port"
Hmm. At least this is partial progress :)
Eric
On 12/7/06, Greg KH <[email protected]> wrote:
> Ugh, no, never use the usb-serial driver as a console device.
>
> That was a bad hack done as a bet many years ago. For many obvious
> reasons it does not work well.
understood, I found with usb_serial convertor could lose some chatacter.
but the usb-debug cable seem it keep all character.
>
> > host with cat /dev/ttyUSB0
> > But if use minicom in host, it will not show '\r', I guess the usb debug
> > cable eat return char. Greg, Can you add that back in usb_debug by
> > replacing '\n' with '\r', '\n'?
>
> The usb-serial console code should handle this, I thought we fixed it a
> while ago.
Is it in the git tree?
>
> But this kind of interface is not what these devices are good for. They
> are for the debug port information, not as a usb-serial console device.
> Otherwise they are way too expensive of a device...
the problem is some "modern" PC will left out serial port. then the cable
could get cheap.
YH
On 12/7/06, Eric W. Biederman <[email protected]> wrote:
> Ugh. I'd check the code. But it looks like my tweak to the
> early fixmap code. But my hunch is that my tweak to __fixmap
> so that it's pud and pmd were prepopulated didn't take on
> your build.
I missed some options?
YH
On 12/8/06, Eric W. Biederman <[email protected]> wrote:
>
> Your or I missed a bug fix/enhancement in there somewhere.
>
I found the problem. the __set_fixmap need to __va, so the entries
will be referred from PAGE_OFFSET.
solution will be
1. move enable_dbgp_console from setup_early_printk, and call it from
setup_arch after init_memory_mapping.
2. or make __set_fixmap can use __pa or pa()+__START_KERNEL_map in
addtion to _va.
please check attached updated patch that is your patch plus 1.
YH
"Yinghai Lu" <[email protected]> writes:
> On 12/8/06, Eric W. Biederman <[email protected]> wrote:
>>
>> Your or I missed a bug fix/enhancement in there somewhere.
>>
>
> I found the problem. the __set_fixmap need to __va, so the entries
> will be referred from PAGE_OFFSET.
>
> solution will be
> 1. move enable_dbgp_console from setup_early_printk, and call it from
> setup_arch after init_memory_mapping.
> 2. or make __set_fixmap can use __pa or pa()+__START_KERNEL_map in
> addtion to _va.
3. Make __va always work. I had this in my tree and I guess it didn't get
into my big rollup patch.
Eric
x86_64: Fix the memory mapping in the early page table
diff --git a/arch/x86_64/kernel/head.S b/arch/x86_64/kernel/head.S
index 1e6f808..2f65469 100644
--- a/arch/x86_64/kernel/head.S
+++ b/arch/x86_64/kernel/head.S
@@ -328,9 +328,9 @@ ENTRY(wakeup_level4_pgt)
.align PAGE_SIZE
ENTRY(boot_level4_pgt)
.quad phys_level3_ident_pgt | 0x007
- .fill 255,8,0
+ .fill 257,8,0
.quad phys_level3_physmem_pgt | 0x007
- .fill 254,8,0
+ .fill 252,8,0
/* (2^48-(2*1024*1024*1024))/(2^39) = 511 */
.quad phys_level3_kernel_pgt | 0x007
On 12/12/06, Eric W. Biederman <[email protected]> wrote:
> diff --git a/arch/x86_64/kernel/head.S b/arch/x86_64/kernel/head.S
> index 1e6f808..2f65469 100644
> --- a/arch/x86_64/kernel/head.S
> +++ b/arch/x86_64/kernel/head.S
> @@ -328,9 +328,9 @@ ENTRY(wakeup_level4_pgt)
> .align PAGE_SIZE
> ENTRY(boot_level4_pgt)
> .quad phys_level3_ident_pgt | 0x007
> - .fill 255,8,0
> + .fill 257,8,0
> .quad phys_level3_physmem_pgt | 0x007
> - .fill 254,8,0
> + .fill 252,8,0
> /* (2^48-(2*1024*1024*1024))/(2^39) = 511 */
> .quad phys_level3_kernel_pgt | 0x007
>
Good, it seems __PAGE_OFFSET used to be 0xfff800000000000
and then 1<<40, and then 0xfff810000000000
YH
wakeup_level4_pgt need to be updated in addtion to boot_level4_pgt?
also comment could be updated for good unstanding too.
YH