Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   "Eric W. Biederman" <ebiederm@xmission.com>
To:     "Li, Xin3" <xin3.li@intel.com>
Cc:     "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "x86@kernel.org" <x86@kernel.org>,
        "tglx@linutronix.de" <tglx@linutronix.de>,
        "mingo@redhat.com" <mingo@redhat.com>,
        "bp@alien8.de" <bp@alien8.de>,
        "dave.hansen@linux.intel.com" <dave.hansen@linux.intel.com>,
        "hpa@zytor.com" <hpa@zytor.com>,
        "brgerst@gmail.com" <brgerst@gmail.com>
References: <20230706052231.2183-1-xin3.li@intel.com>
        <87v8exgmot.fsf@email.froward.int.ebiederm.org>
        <SA1PR11MB67348D11385F584354AB84C3A82DA@SA1PR11MB6734.namprd11.prod.outlook.com>
        <SA1PR11MB6734F3F72095C4D6B9C12C9BA82DA@SA1PR11MB6734.namprd11.prod.outlook.com>
Date:   Fri, 07 Jul 2023 12:28:13 -0500
In-Reply-To: <SA1PR11MB6734F3F72095C4D6B9C12C9BA82DA@SA1PR11MB6734.namprd11.prod.outlook.com>
        (Xin3 Li's message of "Fri, 7 Jul 2023 04:16:47 +0000")
Message-ID: <87sf9zfx76.fsf@email.froward.int.ebiederm.org>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/27.1 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain
Subject: Re: [PATCH] x86/ia32: Do not modify the DPL bits for a null selector
Precedence: bulk

"Li, Xin3" <xin3.li@intel.com> writes:

>> Thus 3 as a selector is the same as 0, and it doesn't matter to change it or not. But
>> when IRET sees an invalid segment register in ES, FS, GS, and DS, it sets it to 0,
>> making 0 a preferred null selector value.
>
> To clarify, an invalid segment register value includes NULL selector values.

Perhaps something like patch below to make it clear that we are
normalizing the segment values and forcing that normalization.

I am a bit confused why this code is not the same for ia32 and
ia32_emulation.  I would think the normalization at least should apply
to the 32bit case as well.

Eric

diff --git a/arch/x86/kernel/signal_32.c b/arch/x86/kernel/signal_32.c
index 9027fc088f97..e5f3978388fd 100644
--- a/arch/x86/kernel/signal_32.c
+++ b/arch/x86/kernel/signal_32.c
@@ -36,22 +36,47 @@
 #ifdef CONFIG_IA32_EMULATION
 #include <asm/ia32_unistd.h>
 
+static inline unsigned int normalize_seg_index(unsigned int index)
+{
+	/*
+	 * Convert the segment index into normalized form.
+	 *
+	 * For the indexes 0,1,2,3 always use the value of 0, as IRET
+	 * forces this form for the nul segment.
+	 *
+	 * Otherwise set both DPL bits to force it to be a userspace
+	 * ring 3 segment index.
+	 */
+	return (index < 3) ? 0 : index | 3;
+}
+
 static inline void reload_segments(struct sigcontext_32 *sc)
 {
-	unsigned int cur;
+	unsigned int new, cur;
 
+	new = normalize_seg_index(sc->gs);
 	savesegment(gs, cur);
-	if ((sc->gs | 0x03) != cur)
-		load_gs_index(sc->gs | 0x03);
+	cur = normalize_seg_index(cur);
+	if (new != cur)
+		load_gs_index(new);
+
+	new = normalize_seg_index(sc->fs);
 	savesegment(fs, cur);
-	if ((sc->fs | 0x03) != cur)
-		loadsegment(fs, sc->fs | 0x03);
+	cur = normalize_seg_index(cur);
+	if (new != cur)
+		loadsegment(fs, new);
+
+	new = normalize_seg_index(sc->ds);
 	savesegment(ds, cur);
-	if ((sc->ds | 0x03) != cur)
-		loadsegment(ds, sc->ds | 0x03);
+	cur = normalize_seg_index(cur);
+	if (new != cur)
+		loadsegment(ds, new);
+
+	new = normalize_seg_index(sc->es);
 	savesegment(es, cur);
-	if ((sc->es | 0x03) != cur)
-		loadsegment(es, sc->es | 0x03);
+	cur = normalize_seg_index(cur);
+	if (new != cur)
+		loadsegment(es, new);
 }
 
 #define sigset32_t			compat_sigset_t