[Crash-utility] [PATCH][v2] x86_64: vtop supports 1G huge pages

[HAGIO KAZUHITO(萩尾 一仁)]

-----Original Message-----
> 
> 在 2020年06月15日 22:18, HAGIO KAZUHITO(萩尾　一仁) 写道:
> > -----Original Message-----
> >> 在 2020年06月12日 00:00, crash-utility-request at redhat.com 写道:
> >>> Looks good to me, thank you Li RongQing.  Please wait for another ack.
> >>
> >> The v2 looks good to me.
> >>
> >> Acked-by: Lianbo Jiang <lijiang at redhat.com>
> >>
> >>>
> >>> Acked-by: Kazuhito Hagio <k-hagio-ab at nec.com>
> >>>
> >>> If someone can comment or create a patch for Xen, please let us know.
> >>> I cannot test it, but I think we can integrate it with this patch if any.
> >>
> >> Anyway, this is a draft patch, but not sure if the patch happens to work
> >> on the Xen. I don't have the Xen environment to confirm it.
> >>
> >> If Xen engineers would like to verify the draft patch, that would be nice.
> >
> > Thank you for your draft patch!
> > If there is no response to the Xen patch until tomorrow morning, I think
> > to apply only Li's patch separately for now.
> >
> OK, sounds good.

OK, Li's patch queued for crash-7.2.9:

https://github.com/crash-utility/crash/commit/c4862e1b2487d0bc7ab4dcfd43d088592a17bece

Thank you all,
Kazu

> 
> > And one more thing, as we discussed internally, we'd like to introduce a
> > title line to crash's commit log for --oneline readability and add helpful
> > information more if available.  On the other hand, I also would like to
> > maintain the crash changelog [0] and its format for searchability.
> >
> Indeed, it makes reading and searching more convenient.
> 
> Thanks.
> Lianbo
> 
> > [0] https://crash-utility.github.io/crash.changelog.html
> >
> > So, for Li's patch, I'll merge it with something like:
> >
> > [commit log]
> > x86_64: Add support for 1GB huge pages to "vtop" command
> >
> > Add support for 1GB huge page to "vtop" command on x86_64.  Without
> > this patch, the command with a user virtual address corresponding to
> > a 1GB huge page fails with the error message "vtop: seek error:
> > physical address: <address>  type: "page table"".
> >
> >   crash> vtop 7f6e40000000
> >   VIRTUAL     PHYSICAL
> >   vtop: seek error: physical address: 3f53f000f000  type: "page table"
> >
> > [crash.changelog]
> >          - Add support for 1GB huge page to "vtop" command on x86_64.  Without
> >            this patch, the command with a user virtual address corresponding to
> >            a 1GB huge page fails with the error message "vtop: seek error:
> >            physical address: <address>  type: "page table"".
> >            (lirongqing at baidu.com, chukaiping at foxmail.com)
> >
> > If you have any problems, please let me know.
> >
> > Thanks,
> > Kazu
> >
> >>
> >> Thanks.
> >> Lianbo
> >>
> >> diff --git a/x86_64.c b/x86_64.c
> >> index fc05e8af5695..90361a7c39c8 100644
> >> --- a/x86_64.c
> >> +++ b/x86_64.c
> >> @@ -2087,6 +2087,7 @@ x86_64_uvtop_level4_xen_wpt(struct task_context *tc, ulong uvaddr, physaddr_t *p
> >>  {
> >>  	ulong pgd_pte;
> >>  	ulong pud_pte;
> >> +	ulong pseudo_pud_pte;
> >>  	ulong pmd_pte;
> >>  	ulong pseudo_pmd_pte;
> >>  	ulong pte;
> >> @@ -2110,6 +2111,38 @@ x86_64_uvtop_level4_xen_wpt(struct task_context *tc, ulong uvaddr, physaddr_t
> *p
> >>  	if (!(pud_pte & _PAGE_PRESENT))
> >>  		goto no_upage;
> >>
> >> +	if (pud_pte & _PAGE_PSE) {
> >> +		if (verbose)
> >> +			fprintf(fp, "  PAGE: %lx  (1GB) [machine]\n",
> >> +				PAGEBASE(pud_pte) & PHYSICAL_PAGE_MASK);
> >> +
> >> +		pseudo_pud_pte = xen_m2p(PAGEBASE(pud_pte));
> >> +
> >> +		if (pseudo_pud_pte == XEN_MACHADDR_NOT_FOUND) {
> >> +			if (verbose)
> >> +				fprintf(fp, " PAGE: page not available\n");
> >> +			*paddr = PADDR_NOT_AVAILABLE;
> >> +			return FALSE;
> >> +		}
> >> +
> >> +		pseudo_pud_pte |= PAGEOFFSET(pud_pte);
> >> +
> >> +		if (verbose) {
> >> +			fprintf(fp, " PAGE: %s  (1GB)\n\n",
> >> +				mkstring(buf, VADDR_PRLEN, RJUST|LONG_HEX,
> >> +				MKSTR(PAGEBASE(pseudo_pud_pte) &
> >> +				PHYSICAL_PAGE_MASK)));
> >> +
> >> +			x86_64_translate_pte(pseudo_pud_pte, 0, 0);
> >> +		}
> >> +
> >> +		physpage = (PAGEBASE(pseudo_pud_pte) & PHYSICAL_PAGE_MASK) +
> >> +			(uvaddr & ~_1GB_PAGE_MASK);
> >> +
> >> +		*paddr = physpage;
> >> +		return TRUE;
> >> +	}
> >> +
> >>  	/*
> >>           *  pmd = pmd_offset(pud, address);
> >>  	 */
> >> @@ -2504,8 +2537,9 @@ no_kpage:
> >>  static int
> >>  x86_64_kvtop_xen_wpt(struct task_context *tc, ulong kvaddr, physaddr_t *paddr, int verbose)
> >>  {
> >> -        ulong *pgd;
> >> +	ulong *pgd;
> >>  	ulong pud_pte;
> >> +	ulong pseudo_pud_pte;
> >>  	ulong pmd_pte;
> >>  	ulong pseudo_pmd_pte;
> >>  	ulong pte;
> >> @@ -2524,6 +2558,36 @@ x86_64_kvtop_xen_wpt(struct task_context *tc, ulong kvaddr, physaddr_t *paddr,
> i
> >>  	if (!(pud_pte & _PAGE_PRESENT))
> >>  		goto no_kpage;
> >>
> >> +	if (pud_pte & _PAGE_PSE) {
> >> +		if (verbose)
> >> +			fprintf(fp, "  PAGE: %lx  (1GB) [machine]\n",
> >> +				PAGEBASE(pud_pte) & PHYSICAL_PAGE_MASK);
> >> +
> >> +		pseudo_pud_pte = xen_m2p(PAGEBASE(pud_pte));
> >> +
> >> +		if (pseudo_pud_pte == XEN_MACHADDR_NOT_FOUND) {
> >> +			if (verbose)
> >> +				fprintf(fp, " PAGE: page not available\n");
> >> +			*paddr = PADDR_NOT_AVAILABLE;
> >> +			return FALSE;
> >> +		}
> >> +
> >> +		pseudo_pud_pte |= PAGEOFFSET(pud_pte);
> >> +		if (verbose) {
> >> +			fprintf(fp, " PAGE: %s  (1GB)\n\n",
> >> +				mkstring(buf, VADDR_PRLEN, RJUST|LONG_HEX,
> >> +				MKSTR(PAGEBASE(pseudo_pud_pte) &
> >> +				PHYSICAL_PAGE_MASK)));
> >> +
> >> +			x86_64_translate_pte(pseudo_pud_pte, 0, 0);
> >> +		}
> >> +		physpage = (PAGEBASE(pseudo_pud_pte) & PHYSICAL_PAGE_MASK) +
> >> +			(kvaddr & ~_1GB_PAGE_MASK);
> >> +
> >> +		*paddr = physpage;
> >> +		return TRUE;
> >> +	}
> >> +
> >>  	/*
> >>  	 *  pmd = pmd_offset(pgd, addr);
> >>  	 */
> >