linux-kernel - Re: [PATCHv5 06/13] x86/mm: Provide ARCH_GET_UNTAG_MASK and ARCH_ENABLE_TAGGED

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: <20220720124744.rpvns3nda7jfljgn@black.fi.intel.com>
Date:   Wed, 20 Jul 2022 15:47:44 +0300
From:   "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>
To:     Alexander Potapenko <glider@...gle.com>
Cc:     Dave Hansen <dave.hansen@...ux.intel.com>,
        Andy Lutomirski <luto@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        the arch/x86 maintainers <x86@...nel.org>,
        Kostya Serebryany <kcc@...gle.com>,
        Andrey Ryabinin <ryabinin.a.a@...il.com>,
        Andrey Konovalov <andreyknvl@...il.com>,
        Taras Madan <tarasmadan@...gle.com>,
        Dmitry Vyukov <dvyukov@...gle.com>,
        "H . J . Lu" <hjl.tools@...il.com>,
        Andi Kleen <ak@...ux.intel.com>,
        Rick Edgecombe <rick.p.edgecombe@...el.com>,
        Linux Memory Management List <linux-mm@...ck.org>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCHv5 06/13] x86/mm: Provide ARCH_GET_UNTAG_MASK and
 ARCH_ENABLE_TAGGED_ADDR

On Wed, Jul 20, 2022 at 10:19:36AM +0200, Alexander Potapenko wrote:
> > > >  long do_arch_prctl_64(struct task_struct *task, int option, unsigned long arg2)
> > > >  {
> > > >         int ret = 0;
> > > > @@ -829,7 +883,11 @@ long do_arch_prctl_64(struct task_struct *task, int option, unsigned long arg2)
> > > >         case ARCH_MAP_VDSO_64:
> > > >                 return prctl_map_vdso(&vdso_image_64, arg2);
> > > >  #endif
> > > > -
> > > > +       case ARCH_GET_UNTAG_MASK:
> > > > +               return put_user(task->mm->context.untag_mask,
> > > > +                               (unsigned long __user *)arg2);
> > >
> > > Can we have ARCH_GET_UNTAG_MASK return the same error value (ENODEV or
> > > EINVAL) as ARCH_ENABLE_TAGGED_ADDR in the case the host doesn't
> > > support LAM?
> > > After all, the mask does not make much sense in this case.
> >
> > I'm not sure about this.
> >
> > As it is ARCH_GET_UNTAG_MASK returns -1UL mask if LAM is not present or
> > not enabled. Applying this mask will give correct result for both.
> 
> Is anyone going to use this mask if tagging is unsupported?
> Tools like HWASan won't even try to proceed in that case.

I can imagine the code that tries to be indifferent to whether a pointer
has tags. It gets mask from ARCH_GET_UNTAG_MASK and applies it to the
pointer without any conditions.

> > Why is -ENODEV better here? Looks like just more work for userspace.
> 
> This boils down to the question of detecting LAM support I raised previously.
> It's nice to have a syscall without side effects to check whether LAM
> can be enabled at all (e.g. one can do the check in the parent process
> and conditionally enable LAM in certain, but not all, child processes)
> CPUID won't help here, because the presence of the LAM bit in CPUID
> doesn't guarantee its support in the kernel, and every other solution
> is more complicated than just issuing a system call.
> 
> Note that TBI has PR_GET_TAGGED_ADDR_CTRL, which can be used to detect
> the presence of memory tagging support.

I would rather make enumeration explicit:

diff --git a/arch/x86/include/uapi/asm/prctl.h b/arch/x86/include/uapi/asm/prctl.h
index 38164a05c23c..a31e27b95b19 100644
--- a/arch/x86/include/uapi/asm/prctl.h
+++ b/arch/x86/include/uapi/asm/prctl.h
@@ -22,5 +22,6 @@

 #define ARCH_GET_UNTAG_MASK		0x4001
 #define ARCH_ENABLE_TAGGED_ADDR		0x4002
+#define ARCH_GET_MAX_TAG_BITS		0x4003

 #endif /* _ASM_X86_PRCTL_H */
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index cfa2e42a135a..2e4df63b775f 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -911,6 +911,13 @@ long do_arch_prctl_64(struct task_struct *task, int option, unsigned long arg2)
 				(unsigned long __user *)arg2);
 	case ARCH_ENABLE_TAGGED_ADDR:
 		return prctl_enable_tagged_addr(task->mm, arg2);
+	case ARCH_GET_MAX_TAG_BITS:
+		if (!cpu_feature_enabled(X86_FEATURE_LAM))
+			return put_user(0, (unsigned long __user *)arg2);
+		else if (lam_u48_allowed())
+			return put_user(15, (unsigned long __user *)arg2);
+		else
+			return put_user(6, (unsigned long __user *)arg2);
 	default:
 		ret = -EINVAL;
 		break;
-- 
 Kirill A. Shutemov