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>] [day] [month] [year] [list]
Message-Id: <20210315111502.440451-1-dwmw2@infradead.org>
Date:   Mon, 15 Mar 2021 11:15:02 +0000
From:   David Woodhouse <dwmw2@...radead.org>
To:     Joerg Roedel <joro@...tes.org>
Cc:     Will Deacon <will@...nel.org>, iommu@...ts.linux-foundation.org,
        linux-kernel@...r.kernel.org, matejm98mthw@...il.com,
        Steven Barrett <steven@...uorix.net>, Mav <basic89@...il.com>
Subject: [PATCH] iommu/amd: Don't initialise remapping irqdomain if IOMMU is disabled

From: David Woodhouse <dwmw@...zon.co.uk>

When the IOMMU is disabled, the driver still enumerates and initialises
the hardware in order to turn it off. Because IRQ remapping setup is
done early, the irqdomain is set up opportunistically.

In commit b34f10c2dc59 ("iommu/amd: Stop irq_remapping_select() matching
when remapping is disabled") I already make the irq_remapping_select()
function check the amd_iommu_irq_setup flag because that might get
cleared only after the irqdomain setup is done, when the IVRS is parsed.

However, in the case where 'amd_iommu=off' is passed on the command line,
the IRQ remapping setup isn't done but the amd_iommu_irq_setup flag is
still set by the early IRQ remap init code. Stop it doing that, by
bailing out of amd_iommu_prepare() early when it's disabled.

This avoids the crash in irq_remapping_select() as it dereferences the
NULL amd_iommu_rlookup_table[]:

[    0.243659] Switched APIC routing to physical x2apic.
[    0.262206] BUG: kernel NULL pointer dereference, address: 0000000000000500
[    0.262927] #PF: supervisor read access in kernel mode
[    0.263390] #PF: error_code(0x0000) - not-present page
[    0.263844] PGD 0 P4D 0
[    0.264135] Oops: 0000 [#1] SMP PTI
[    0.264460] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.12.0-rc3 #831
[    0.265069] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.14.0-1.fc33 04/01/2014
[    0.265825] RIP: 0010:irq_remapping_select+0x57/0xb0
[    0.266327] Code: 4b 0c 48 3d 30 e0 a7 9e 75 0d eb 35 48 8b 00 48 3d 30 e0 a7 9e 74 2a 0f b6 50 10 39 d1 75 ed 0f b7 40 12 48 8b 15 69 e3 d2 01 <48> 8b 14 c2 48 85 d2 74 0e b8 01 00 00 00 48 3b aa 90 04 00 00 74
[    0.268412] RSP: 0000:ffffffff9e803db0 EFLAGS: 00010246
[    0.268919] RAX: 00000000000000a0 RBX: ffffffff9e803df8 RCX: 0000000000000000
[    0.269550] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff98120112fe79
[    0.270245] RBP: ffff9812011c8218 R08: 0000000000000001 R09: 000000000000000a
[    0.270922] R10: 000000000000000a R11: f000000000000000 R12: ffff9812011c8218
[    0.271549] R13: ffff98120181ed88 R14: 0000000000000000 R15: 0000000000000000
[    0.272221] FS:  0000000000000000(0000) GS:ffff98127dc00000(0000) knlGS:0000000000000000
[    0.272997] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    0.273508] CR2: 0000000000000500 CR3: 0000000030810000 CR4: 00000000000006b0
[    0.274178] Call Trace:
[    0.274416]  irq_find_matching_fwspec+0x41/0xc0
[    0.274812]  mp_irqdomain_create+0x65/0x150
[    0.275251]  setup_IO_APIC+0x70/0x811

Fixes: a1a785b57242 ("iommu/amd: Implement select() method on remapping irqdomain")
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=212017
Signed-off-by: David Woodhouse <dwmw@...zon.co.uk>
---
 drivers/iommu/amd/init.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/iommu/amd/init.c b/drivers/iommu/amd/init.c
index 9126efcbaf2c..07edd837b076 100644
--- a/drivers/iommu/amd/init.c
+++ b/drivers/iommu/amd/init.c
@@ -2998,6 +2998,9 @@ int __init amd_iommu_prepare(void)
 {
 	int ret;
 
+	if (amd_iommu_disabled)
+		return -ENODEV;
+
 	amd_iommu_irq_remap = true;
 
 	ret = iommu_go_to_state(IOMMU_ACPI_FINISHED);
-- 
2.29.2

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ