Re: [PATCH v2] x86_64, vmcoreinfo: Append 'page_offset_base' to vmcoreinfo

[Borislav Petkov]

+ Kees.

On Fri, Nov 16, 2018 at 03:17:49AM +0530, Bhupesh Sharma wrote:
> x86_64 kernel uses 'page_offset_base' variable to point to the
> start of direct mapping of all physical memory. This variable
> is also updated for KASLR boot cases, so this can be exported
> via vmcoreinfo as a standard ABI between kernel and user-space,
> to allow user-space utilities to use the same for calculating
> the start of direct mapping of all physical memory.
> 
> 'arch/x86/kernel/head64.c' sets the same as:
>    unsigned long page_offset_base __ro_after_init = __PAGE_OFFSET_BASE_L4;
> 
> and also uses the same to indicate the base of KASLR regions on x86_64:
>    static __initdata struct kaslr_memory_region {
> 	    unsigned long *base;
> 	        unsigned long size_tb;
>    } kaslr_regions[] = {
> 	    { &page_offset_base, 0 },
>    .. snip ..

Why is that detail needed in the commit message?

> Adding 'page_offset_base' to the vmcoreinfo can be specially useful for
> live-debugging of a running kernel via user-space utilities
> like makedumpfile (see [1]).
> 
> Recently, I saw an issue with the 'makedumpfile' utility (see [2] for

Use passive tone in your commit message: no "we" or "I", etc.

Also, pls read section "2) Describe your changes" in
Documentation/process/submitting-patches.rst.

> details), whose live debugging feature is broken with newer kernels
> (I tested the same with 4.19-rc8+ kernel), as KCORE_REMAP segments were
> added to kcore, thus leading to an additional sections in the same, and
> makedumpfile is not longer able to determine the start of direct
> mapping of all physical memory, as it relies on traversing the PT_LOAD
> segments inside kcore and using the last PT_LOAD segment
> to determine the start of direct mapping.
> 
> Such user-space issues can be resolved if the user-space code instead
> uses a standard ABI to read the kernel exposed machine specific
> variables. With the kernel commit 23c85094fe1895caefdd
> ["proc/kcore: add vmcoreinfo note to /proc/kcore"]), it is

ERROR: Please use git commit description style 'commit <12+ chars of sha1> ("<title line>")' - ie: 'commit 23c85094fe18 ("proc/kcore: add vmcoreinfo note to /proc/kcore")'
#54: 
variables. With the kernel commit 23c85094fe1895caefdd

> now possible to use the vmcoreinfo present inside kcore as the standard
> ABI which can be used by the user-space utilities for reading
> the machine specific information (and hence for debugging a
> live kernel).
> 
> User-space utilities like makedumpfile, kexec-tools and crash
> are either already using this ABI or are discussing patches
> which look to add the same feature. This helps in simplifying the
> overall code and also in reducing code-rewrite across the
> user-space utilities for getting values of these kernel
> symbols/variables.

> Accordingly this patch allows appending 'page_offset_base' for
> x86_64 platforms to vmcoreinfo, so that user-space tools can use the
> same as a standard interface to determine the start of direct mapping
> of all physical memory.
> 
> Testing:
> -------
>  - I tested this patch (rebased on 'linux-next') on a x86_64 machine
>    using the modified 'makedumpfile' user-space code (see [3] for my
>    github tree which contains the same) for determining how many pages
>    are dumpable when different dump_level is specified (which is
>    one use-case of live-debugging via 'makedumpfile').
>  - I tested both the KASLR and non-KASLR boot cases with this patch.
>  - Here is one sample log (for KASLR boot case) on my x86_64 machine:
> 
>    < snip..>
>    The kernel doesn't support mmap(),read() will be used instead.
> 
>    TYPE		PAGES			EXCLUDABLE	DESCRIPTION
>    ----------------------------------------------------------------------
>    ZERO		21299           	yes		Pages filled
>    with zero
>    NON_PRI_CACHE	91785           	yes		Cache
>    pages without private flag
>    PRI_CACHE	1               	yes		Cache pages with
>    private flag
>    USER		14057           	yes		User process
>    pages
>    FREE		740346          	yes		Free pages
>    KERN_DATA	58152           	no		Dumpable kernel
>    data
> 
>    page size:		4096
>    Total pages on system:	925640
>    Total size on system:	3791421440       Byte
> 
> [1]. MAN pages -> MAKEDUMPFILE(8) and CRASH(8)
> [2]. makedumpfile issue with latest kernels -> http://lists.infradead.org/pipermail/kexec/2018-October/021769.html
> [3]. https://github.com/bhupesh-sharma/makedumpfile/tree/add-page-offset-base-to-vmcore-v1
> 
> Cc: Boris Petkov <bp@xxxxxxxxx>
> Cc: Baoquan He <bhe@xxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: Kazuhito Hagio <k-hagio@xxxxxxxxxxxxx>
> Cc: Dave Anderson <anderson@xxxxxxxxxx>
> Cc: James Morse <james.morse@xxxxxxx>
> Cc: Omar Sandoval <osandov@xxxxxx>
> Cc: x86@xxxxxxxxxx
> Cc: kexec@xxxxxxxxxxxxxxxxxxx
> Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
> Signed-off-by: Bhupesh Sharma <bhsharma@xxxxxxxxxx>
> ---
> Changes since v1:
>  - Fixed the build issue reported by build bot and tested this version
>    with 'make allmodconfig'.
>  - Reworded most of the commit log to explain the intent behind the
>    patch.
> 
>  arch/x86/kernel/machine_kexec_64.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/arch/x86/kernel/machine_kexec_64.c b/arch/x86/kernel/machine_kexec_64.c
> index 4c8acdfdc5a7..6161d77c5bfb 100644
> --- a/arch/x86/kernel/machine_kexec_64.c
> +++ b/arch/x86/kernel/machine_kexec_64.c
> @@ -356,6 +356,9 @@ void arch_crash_save_vmcoreinfo(void)
>  	VMCOREINFO_SYMBOL(init_top_pgt);
>  	vmcoreinfo_append_str("NUMBER(pgtable_l5_enabled)=%d\n",
>  			pgtable_l5_enabled());
> +#ifdef CONFIG_RANDOMIZE_BASE
> +	VMCOREINFO_NUMBER(page_offset_base);
> +#endif
>  
>  #ifdef CONFIG_NUMA
>  	VMCOREINFO_SYMBOL(node_data);
> --

All above are only nitpicks though.

My opinion is this: people are exporting all kinds of kernel-internal
stuff in vmcoreinfo and frankly, I'm not crazy about this idea.

And AFAICT, this thing basically bypasses KASLR completely but you need
root for it so it probably doesn't really matter.

Now, on another thread we agreed more or less that what gets exported in
vmcoreinfo is so tightly coupled to the running kernel so that it is not
even considered an ABI. I guess that is debatable but whatever.

So my only request right now would be to have all those things being
exported, documented somewhere and I believe Lianbo is working on that.

But I'm sure others will have more to say about it.

-- 
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.