(adding Hans, Kirill and Ingo) Exec summary: youling is seeing issues on his Bay Trail tablet booting the 64-bit kernel from 32-bit EFI (aka mixed mode) He has narrowed it down to the patch below. Unfortunately, I cannot reproduce this on QEMU/kvm running in mixed mode, and Hans has not reported anything similar either. I am reasonably familiar now with the way the thunking between 64-bit and 32-bit is implemented, but I am clueless about how x86 configures the page tables etc so we need some help from team-x86 figuring this out. Thanks, Ard. On 16 July 2018 at 22:13, youling 257 <youling257@xxxxxxxxx> wrote: > Final，i found the problem patch is > 32fcefa2bfc8961987e91d1daeb00624b4176d2e，x86/boot/compressed/64: Set up > trampoline memory > Youling, Thanks a lot for narrowing that down! We will try to investigate in more detail. > 2018-07-16 15:30 GMT+08:00 Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>: >> >> >> >> On 16 Jul 2018, at 06:00, youling 257 <youling257@xxxxxxxxx> wrote: >> >> >> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/log/?h=x86-mm-for-linus >> ，same problem，it's bad for me. >> git checkout eaeb8e76cd5751e805f6e4a3fcec91d283e3b0c2 >> git format-patch -s 3eb2ce825ea1ad89d20f7a3b5780df850e4be274 >> will get extra 56 patches，4.16rc7 is good for me. >> >> of course，between 986b37c0ae4f0a3f93d8974d03a9cbc1502dd377 and >> d22fff81418edc92be534cad8d59da914049bf69 is 56 extra patches. >> >> now，I don't know how to find problem reason. continue bisect not a nice >> method. >> >> >> Why not? If you have a good and a bad commit and 56 possible candidates, >> bisect should help you find the bad commit very quickly >> >> >> 2018-07-16 4:56 GMT+08:00 Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>: >>> >>> (+ Sai) >>> >>> Please don't remove cc's - I added them for a reason. >>> >>> On 15 July 2018 at 11:14, youling 257 <youling257@xxxxxxxxx> wrote: >>> > build kernel many times，uname -a，#1，#2，#3，#4，#5，#6，#7，#8，#9，ok ? >>> >>> I see. Unfortunately, those numbers are absolutely meaningless when it >>> comes to describing the exact configuration they were built from. >>> >>> > only #9 kernel config is CONFIG_X86_5LEVEL=y，#1 kernel config >>> > CONFIG_X86_5LEVEL is not set . >>> > can you understand >>> > >>> >>> OK. >>> >>> The merge commit you identified makes a couple of x86/mm related changes. >>> >>> That is why I asked you about this configuration. >>> >>> >>> > 2018-07-15 17:11 GMT+08:00 Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>: >>> >> >>> >> >>> >> >>> >> On 15 Jul 2018, at 10:57, youling 257 <youling257@xxxxxxxxx> wrote: >>> >> >>> >> I had try it，my #9 kernel CONFIG_X86_5LEVEL=y，same problem. other #1 >>> >> to >>> >> #8 not set 5level. >>> >> so you say must CONFIG_X86_5LEVEL= y ? >>> >> >>> >> >>> >> I don’t know what #1 to #9 means, but the first failing commit does >>> >> enable >>> >> 5 level paging in generic builds >>> >> >>> >> please try disabling it altogether >>> >> >>> >> 2018-07-15 16:42 GMT+08:00 Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>: >>> >>> >>> >>> On 14 July 2018 at 19:36, youling 257 <youling257@xxxxxxxxx> wrote: >>> >>> > Bisect branch，build and test boot，last good branch is >>> >>> > 986b37c0ae4f0a3f93d8974d03a9cbc1502dd377，first bad branch is >>> >>> > d22fff81418edc92be534cad8d59da914049bf69 . >>> >>> > >>> >>> > >>> >>> > https://github.com/torvalds/linux/commits/2fcd2b306aa80771e053275ed74b2dfe7e3d1434 >>> >>> > >>> >>> >>> >>> Does you kernel config have CONFIG_X86_5LEVEL=y set? >>> >>> >>> >>> >>> >>> > youling 257 <youling257@xxxxxxxxx> 于 2018年7月14日周六 上午8:15写道： >>> >>> >> >>> >>> >> just like these，i only can copy these from webpage，i can't use >>> >>> >> English >>> >>> >> say >>> >>> >> clear my Bay trail 64 bit 4.17 kernel on 32 bit uefi problem. 4.17 >>> >>> >> kernel，sometimes，kept rebooting itself during loading of the >>> >>> >> kernel >>> >>> >> and >>> >>> >> would be a never ending loop of reboots. 4.17 kernel，sometimes >>> >>> >> hang on >>> >>> >> reboot when using that kernel version. It goes to bios screen and >>> >>> >> hangs for >>> >>> >> like 1-2 min and then a blank screen with a _ blinking. I could >>> >>> >> then >>> >>> >> shutoff >>> >>> >> my computer using the power button.Shutting down is fine though. >>> >>> >> >>> >>> >> Prakhya, Sai Praneeth <sai.praneeth.prakhya@xxxxxxxxx> 于 >>> >>> >> 2018年7月14日周六 >>> >>> >> 上午5:18写道： >>> >>> >>> >>> >>> >>> OK.. thanks for this info. Could you please answer *all* my below >>> >>> >>> questions (not just the git bisect one). I didn’t understand the >>> >>> >>> problem in >>> >>> >>> the first place. >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> I build and test 4.17 rc1 kernel，without doubt，the first problem >>> >>> >>> kernel >>> >>> >>> version is 4.17 rc1. >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> Could you please describe the problem you are seeing again? >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> 1. What are the symptoms? Does the system not boot at all or >>> >>> >>> is >>> >>> >>> it >>> >>> >>> hanging during reboot? Do you see a stack trace? If so, please >>> >>> >>> show >>> >>> >>> that, >>> >>> >>> that will be *very helpful* to debug. >>> >>> >>> >>> >>> >>> 2. Effected system? (Baytrail or Toshiba Satellite). Please >>> >>> >>> understand that we might not have access to all the machines, so >>> >>> >>> it >>> >>> >>> would be >>> >>> >>> easier for us to fix the issue if it’s widespread but we will >>> >>> >>> still >>> >>> >>> do our >>> >>> >>> best to fix all issues. >>> >>> >>> >>> >>> >>> 3. What’s your config file and kernel command line? >>> >>> >>> >>> >>> >>> 4. What’s the last working kernel version and what’s the >>> >>> >>> first >>> >>> >>> broken kernel version? (Ideally the difference should only be one >>> >>> >>> version >>> >>> >>> (Eg: v4.17-rc3 working and v4.17-rc4 not working). As Ard >>> >>> >>> suggested, >>> >>> >>> we >>> >>> >>> would highly appreciate if you could track down the issue to a >>> >>> >>> single >>> >>> >>> patch >>> >>> >>> using git bisect [1]. >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> [1] A quick google search on git bisect will teach you how to use >>> >>> >>> it >>> >>> >>> or >>> >>> >>> if you find it hard, let me know, I can help you with that. >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> Regards, >>> >>> >>> >>> >>> >>> Sai >>> >>> >>> >>> >>> >>> >>> >> >>> >> >>> > >> >> > -- To unsubscribe from this list: send the line "unsubscribe linux-efi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
