Re: [LKP] [x86/KASLR] ed9f007ee6: -- System halted

[Baoquan He]

On 07/01/16 at 12:06pm, Baoquan He wrote:
> On 06/30/16 at 05:24pm, Yinghai Lu wrote:
> > On Wed, Jun 29, 2016 at 11:58 PM, Ye Xiaolong <xiaolong.ye@xxxxxxxxx> wrote:
> > > On Wed, Jun 29, 2016 at 09:49:12PM -0700, Yinghai Lu wrote:
> > >>On Wed, Jun 29, 2016 at 1:29 PM, Huang, Ying <ying.huang@xxxxxxxxx> wrote:
> > >>> If you could provide a git branch for that, that will be easier for us
> > >>> to test and more accurate for you to get the right patch to be tested.
> > >>>
> > >>
> > >>Please check
> > >>
> > >>git://git.kernel.org/pub/scm/linux/kernel/git/yinghai/linux-yinghai.git
> > >>for-x86-v4.8-kaslr-debug
> > >
> > > Hi, yinghai
> > >
> > > Attached is the dmesg for the kernel which is built based on your
> > > for-x86-v4.8-kaslr-debug branch.
> > 
> > Thanks for capturing the boot log.
> > 
> > early console in setup code
> > early console in extract_kernel
> > decompress_kernel:
> >        input: [0x0008fba276-0x000a8818ed]
> >       output: [0x0001000000-0x000a867c97] 0x09867c98: output_len
> >               [0x0001000000-0x0009768fff] 0x08769000: run_size
> >               [0x0001000000-0x000a867c97] 0x09867c98: output_run_size
> >               [0x0001000000-0x000a8affff] 0x098b0000: init_size
> > ZO text/data: [0x000a8818ee-0x000a8affff]
> >      ZO heap: [0x000a888400-0x000a8983ff]
> >   VO bss/brk: [0x0009674000-0x0009768fff]
> > booted via startup_32()
> > Physical KASLR using RDTSC...
> > Virtual KASLR using RDTSC...
> >   new output: [0x0006600000-0x000fe67c97] 0x09867c98: output_run_size
> >    virt_addr:  0x0024000000
> >   decompress: [0x0006600000-0x000fe67c97] <=== [0x0008fba276-0x000a8818ed]
> > 
> > Decompressing Linux...
> > 
> > XZ-compressed data is corrupt
> > 
> >  -- System halted
> > 
> > Hi Kees/Baoquan,
> > 
> > Looks like
> > 
> > commit ed9f007ee68478f6a50ec9971ade25a0129a5c0e
> > Author: Kees Cook <keescook@xxxxxxxxxxxx>
> > Date:   Wed May 25 15:45:33 2016 -0700
> > 
> >     x86/KASLR: Extend kernel image physical address randomization to addresses l
> > arger than 4G
> 
> I know why it happened, making patch. will reply soon.

Sorry, I didn't get it, continue checking.

> 
> > 
> > 
> > cause choose_random_location() return new output buffer
> > new output: [0x0006600000-0x000fe67c97] 0x09867c98: output_run_size
> > 
> > that is overlapped with input data.
> >        input: [0x0008fba276-0x000a8818ed]
> >               ....-0x000a8affff] 0x098b0000: init_size
> > 
> > and input already is in mem_avoid.