From:   Juergen Gross <jgross@suse.com>
To:     linux-kernel@vger.kernel.org, x86@kernel.org
Cc:     lists@nerdbynature.de, mikelley@microsoft.com,
        torvalds@linux-foundation.org, Juergen Gross <jgross@suse.com>,
        Boris Ostrovsky <boris.ostrovsky@oracle.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        "H. Peter Anvin" <hpa@zytor.com>, xen-devel@lists.xenproject.org
Subject: [PATCH v2 4/8] x86/xen: set MTRR state when running as Xen PV initial domain
Date:   Thu,  9 Feb 2023 08:22:16 +0100
Message-Id: <20230209072220.6836-5-jgross@suse.com>
In-Reply-To: <20230209072220.6836-1-jgross@suse.com>
References: <20230209072220.6836-1-jgross@suse.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

When running as Xen PV initial domain (aka dom0), MTRRs are disabled
by the hypervisor, but the system should nevertheless use correct
cache memory types. This has always kind of worked, as disabled MTRRs
resulted in disabled PAT, too, so that the kernel avoided code paths
resulting in inconsistencies. This bypassed all of the sanity checks
the kernel is doing with enabled MTRRs in order to avoid memory
mappings with conflicting memory types.

This has been changed recently, leading to PAT being accepted to be
enabled, while MTRRs stayed disabled. The result is that
mtrr_type_lookup() no longer is accepting all memory type requests,
but started to return WB even if UC- was requested. This led to
driver failures during initialization of some devices.

In reality MTRRs are still in effect, but they are under complete
control of the Xen hypervisor. It is possible, however, to retrieve
the MTRR settings from the hypervisor.

In order to fix those problems, overwrite the MTRR state via
mtrr_overwrite_state() with the MTRR data from the hypervisor, if the
system is running as a Xen dom0.

Fixes: 72cbc8f04fe2 ("x86/PAT: Have pat_enabled() properly reflect state when running on Xen")
Signed-off-by: Juergen Gross <jgross@suse.com>
---
V2:
- new patch
---
 arch/x86/xen/enlighten_pv.c | 49 +++++++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

diff --git a/arch/x86/xen/enlighten_pv.c b/arch/x86/xen/enlighten_pv.c
index 5b1379662877..9cf520c0c810 100644
--- a/arch/x86/xen/enlighten_pv.c
+++ b/arch/x86/xen/enlighten_pv.c
@@ -68,6 +68,7 @@
 #include <asm/reboot.h>
 #include <asm/hypervisor.h>
 #include <asm/mach_traps.h>
+#include <asm/mtrr.h>
 #include <asm/mwait.h>
 #include <asm/pci_x86.h>
 #include <asm/cpu.h>
@@ -1200,6 +1201,52 @@ static void __init xen_boot_params_init_edd(void)
 #endif
 }
 
+/* Get MTRR settings from Xen and put them into mtrr_state. */
+static void __init xen_set_mtrr_data(void)
+{
+#ifdef CONFIG_MTRR
+	struct xen_platform_op op = {
+		.cmd = XENPF_read_memtype,
+		.interface_version = XENPF_INTERFACE_VERSION,
+	};
+	unsigned int reg;
+	unsigned long mask;
+	uint32_t eax, width;
+	static struct mtrr_var_range var[MTRR_MAX_VAR_RANGES] __initdata;
+
+	/* Get physical address width (only 64-bit cpus supported). */
+	width = 36;
+	eax = cpuid_eax(0x80000000);
+	if ((eax >> 16) == 0x8000 && eax >= 0x80000008) {
+		eax = cpuid_eax(0x80000008);
+		width = eax & 0xff;
+	}
+
+	for (reg = 0; reg < MTRR_MAX_VAR_RANGES; reg++) {
+		op.u.read_memtype.reg = reg;
+		if (HYPERVISOR_platform_op(&op))
+			break;
+
+		/*
+		 * Only called in dom0, which has all RAM PFNs mapped at
+		 * RAM MFNs, and all PCI space etc. is identity mapped.
+		 * This means we can treat MFN == PFN regarding MTTR settings.
+		 */
+		var[reg].base_lo = op.u.read_memtype.type;
+		var[reg].base_lo |= op.u.read_memtype.mfn << PAGE_SHIFT;
+		var[reg].base_hi = op.u.read_memtype.mfn >> (32 - PAGE_SHIFT);
+		mask = ~((op.u.read_memtype.nr_mfns << PAGE_SHIFT) - 1);
+		mask &= (1UL << width) - 1;
+		if (mask)
+			mask |= 1 << 11;
+		var[reg].mask_lo = mask;
+		var[reg].mask_hi = mask >> 32;
+	}
+
+	mtrr_overwrite_state(var, reg, NULL, MTRR_TYPE_UNCACHABLE);
+#endif
+}
+
 /*
  * Set up the GDT and segment registers for -fstack-protector.  Until
  * we do this, we have to be careful not to call any stack-protected
@@ -1403,6 +1450,8 @@ asmlinkage __visible void __init xen_start_kernel(struct start_info *si)
 
 		xen_boot_params_init_edd();
 
+		xen_set_mtrr_data();
+
 #ifdef CONFIG_ACPI
 		/*
 		 * Disable selecting "Firmware First mode" for correctable
-- 
2.35.3