linux-kernel - Re: [PATCH 0/5] [v3] fix illegal use of _

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20130225083039.GB21422@redhat.com>
Date:	Mon, 25 Feb 2013 10:30:39 +0200
From:	Gleb Natapov <gleb@...hat.com>
To:	Peter Hurley <peter@...leysoftware.com>
Cc:	Dave Hansen <dave@...ux.vnet.ibm.com>,
	linux-kernel@...r.kernel.org, linux-mm@...ck.org,
	"H. Peter Anvin" <hpa@...or.com>, x86@...nel.org,
	Marcelo Tosatti <mtosatti@...hat.com>,
	Rik van Riel <riel@...hat.com>
Subject: Re: [PATCH 0/5] [v3] fix illegal use of __pa() in KVM code

On Sun, Feb 24, 2013 at 04:28:58PM -0500, Peter Hurley wrote:
> 
> On Tue, 2013-01-22 at 13:24 -0800, Dave Hansen wrote:
> > This series fixes a hard-to-debug early boot hang on 32-bit
> > NUMA systems.  It adds coverage to the debugging code,
> > adds some helpers, and eventually fixes the original bug I
> > was hitting.
> 
> Hi Dave,
> 
> Now that the alloc_remap() has been/is being removed, is most/all of
> this being reverted?
> 
> I ask because I was fixing a different bug in KVM's para-virt clock and
> saw part of this series (from [PATCH 5/5] fix kvm's use of __pa() on
> percpu areas):
> 
> diff -puN arch/x86/kernel/kvmclock.c~fix-kvm-__pa-use-on-percpu-areas arch/x86/kernel/kvmclock.c
> --- linux-2.6.git/arch/x86/kernel/kvmclock.c~fix-kvm-__pa-use-on-percpu-areas   2013-01-22 13:17:16.428317508 -0800
> +++ linux-2.6.git-dave/arch/x86/kernel/kvmclock.c       2013-01-22 13:17:16.432317541 -0800
> @@ -162,8 +162,8 @@ int kvm_register_clock(char *txt)
>         int low, high, ret;
>         struct pvclock_vcpu_time_info *src = &hv_clock[cpu].pvti;
>  
> -       low = (int)__pa(src) | 1;
> -       high = ((u64)__pa(src) >> 32);
> +       low = (int)slow_virt_to_phys(src) | 1;
> +       high = ((u64)slow_virt_to_phys(src) >> 32);
>         ret = native_write_msr_safe(msr_kvm_system_time, low, high);
>         printk(KERN_INFO "kvm-clock: cpu %d, msr %x:%x, %s\n",
>                cpu, high, low, txt);
> 
I think this hunk was converted by mistake since the code does not use
per cpu allocations, but since __pa() and slow_virt_to_phys() should
return the same thing here it does not matter much.

> which confused me because hv_clock is the __va of allocated physical
> memory, not a per-cpu variable.
> 
> 	mem = memblock_alloc(size, PAGE_SIZE);
> 	if (!mem)
> 		return;
> 	hv_clock = __va(mem);
> 
> So in short, my questions are:
> 1) is the slow_virt_to_phys() necessary anymore?
> 2) if yes, does it apply to the code above?
> 3) if yes, would you explain in more detail what the 32-bit NUMA mm is
> doing, esp. wrt. when __va(__pa) is not identical across all cpus?
> 
> Regards,
> Peter Hurley
> 

--
			Gleb.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/