Messages in this thread | | | Date | Sun, 1 Mar 2015 11:27:48 -0800 | Subject | Re: [PATCH 1/8] x86, kaslr: get kaslr_enabled back correctly | From | Yinghai Lu <> |
| |
On Sun, Mar 1, 2015 at 7:23 AM, Ingo Molnar <mingo@kernel.org> wrote: > > I think that's a different bug. > > parse_kaslr_setup() is simply bogus, it does: > > kaslr_enabled = (bool)(pa_data + sizeof(struct setup_data)); > > which makes no sense whatsoever: it randomly enables (or disables, > depending on the physical address of the setup page) KASLR when it > meets a SETUP_KASLR record. >
Yes, you are right.
there are 8 patches in this series. x86, kaslr: get kaslr_enabled back correctly x86: Kill E820_RESERVED_KERN x86, efi: copy SETUP_EFI data and access directly x86, of: let add_dtb reserve by itself x86, boot: Add add_pci handler for SETUP_PCI x86: kill not used setup_data handling code x86, pci: convert SETUP_PCI data to list x86, pci: export SETUP_PCI data via sysfs
other 7 should also address the problem in http://lkml.kernel.org/r/1424929021.10337.24.camel@intel.com
the root cause is E820_RESERVED_KERN and setup_data handling will make E820 are not page aligned anymore.
We should change setup_data handling.
Please check updated version at
git://git.kernel.org/pub/scm/linux/kernel/git/yinghai/linux-yinghai.git for-x86-4.0
(with a fix for config without CONFIG_PCI).
Thanks
Yinghai
| |