Re: [PATCH] irq-gic-v3: fix NULL dereference of disabled redist_base

From: Marc Zyngier
Date: Mon Dec 16 2019 - 06:14:58 EST

Hi Heyi,

On 2019-12-16 06:27, Heyi Guo wrote:
If we use ACPI MADT GICC structure to pass single redistributor base,
and mark some GICC as disabled, we'll get below call trace during
boot:

[ 0.000000] NR_IRQS: 64, nr_irqs: 64, preallocated irqs: 0
[ 0.000000] GICv3: 256 SPIs implemented
[ 0.000000] GICv3: 0 Extended SPIs implemented
[ 0.000000] GICv3: Distributor has no Range Selector support
[ 0.000000] Unable to handle kernel paging request at virtual
address 000000000000ffe8
[ 0.000000] Mem abort info:
[ 0.000000] ESR = 0x96000004
[ 0.000000] EC = 0x25: DABT (current EL), IL = 32 bits
[ 0.000000] SET = 0, FnV = 0
[ 0.000000] EA = 0, S1PTW = 0
[ 0.000000] Data abort info:
[ 0.000000] ISV = 0, ISS = 0x00000004
[ 0.000000] CM = 0, WnR = 0
[ 0.000000] [000000000000ffe8] user address but active_mm is swapper
[ 0.000000] Internal error: Oops: 96000004 [#1] SMP
[ 0.000000] Modules linked in:
[ 0.000000] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.5.0-rc1 #5
[ 0.000000] pstate: 20000085 (nzCv daIf -PAN -UAO)
[ 0.000000] pc : gic_iterate_rdists+0x58/0x130
[ 0.000000] lr : gic_iterate_rdists+0x80/0x130
[ 0.000000] sp : ffff8000113d3cb0
[ 0.000000] x29: ffff8000113d3cb0 x28: 0000000000000000
[ 0.000000] x27: 0000000000000000 x26: 0000000000000018
[ 0.000000] x25: 000000000000ffe8 x24: 000000000000003f
[ 0.000000] x23: ffff800010588040 x22: 00000000000005e8
[ 0.000000] x21: ffff8000113df7d0 x20: 0000030f00003f11
[ 0.000000] x19: 0000000000000000 x18: ffffffffffffffff
[ 0.000000] x17: 0000000014aeb8dc x16: 00000000c3ba0ccf
[ 0.000000] x15: ffff8000113d9908 x14: ffff8000913d3a37
[ 0.000000] x13: ffff8000113d3a45 x12: ffff800011402000
[ 0.000000] x11: ffff8000113d39d0 x10: ffff8000113db980
[ 0.000000] x9 : 00000000ffffffd0 x8 : ffff8000106dca98
[ 0.000000] x7 : 000000000000005b x6 : 0000000000000000
[ 0.000000] x5 : 0000000000000000 x4 : ffff8000128c0000
[ 0.000000] x3 : ffff8000128a0000 x2 : ffff0003fc3c7000
[ 0.000000] x1 : 0000000000000001 x0 : 000000000000ffe8
[ 0.000000] Call trace:
[ 0.000000] gic_iterate_rdists+0x58/0x130
[ 0.000000] gic_init_bases+0x200/0x4b4
[ 0.000000] gic_acpi_init+0x148/0x284
[ 0.000000] acpi_match_madt+0x4c/0x84
[ 0.000000] acpi_table_parse_entries_array+0x188/0x278
[ 0.000000] acpi_table_parse_entries+0x70/0x98
[ 0.000000] acpi_table_parse_madt+0x40/0x50
[ 0.000000] __acpi_probe_device_table+0x88/0xe4
[ 0.000000] irqchip_init+0x38/0x40
[ 0.000000] init_IRQ+0x168/0x19c
[ 0.000000] start_kernel+0x328/0x508
[ 0.000000] Code: f90017b6 9b3a7f16 f8766853 8b190260 (b9400000)
[ 0.000000] ---[ end trace ae5cf232d924bfc1 ]---
[ 0.000000] Kernel panic - not syncing: Fatal exception
[ 0.000000] Rebooting in 3 seconds..

In this case, nr_redist_regions counts all GICC structures but only
enabled ones have redistributor mapped. So add check to avoid NULL
deference of redist_base.

Signed-off-by: Heyi Guo <guoheyi@xxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Jason Cooper <jason@xxxxxxxxxxxxxx>
Cc: Marc Zyngier <maz@xxxxxxxxxx>
---
drivers/irqchip/irq-gic-v3.c | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c
index d6218012097b..bd9d55cadef9 100644
--- a/drivers/irqchip/irq-gic-v3.c
+++ b/drivers/irqchip/irq-gic-v3.c
@@ -781,6 +781,13 @@ static int gic_iterate_rdists(int (*fn)(struct
redist_region *, void __iomem *))
u64 typer;
u32 reg;

+ /*
+ * redist_base may be NULL if we use single_redist and some GICC
+ * structure is disabled.
+ */
+ if (!ptr)
+ continue;
+
reg = readl_relaxed(ptr + GICR_PIDR2) & GIC_PIDR2_ARCH_MASK;
if (reg != GIC_PIDR2_ARCH_GICv3 &&
reg != GIC_PIDR2_ARCH_GICv4) { /* We're in trouble... */

This feels like the wrong fix. The redistributor region array should
be completely populated, and there is an assumption all over this driver
that there is no junk in these structures.

You're seeing this because we don't track the number of *enabled* rdists,
and allocate the number of regions based on the number of overall GICC
entries instead of the number of enabled redistributors.

How about this instead?

M.

diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c
index 3a1866682dd0..9b26a860340b 100644
--- a/drivers/irqchip/irq-gic-v3.c
+++ b/drivers/irqchip/irq-gic-v3.c
@@ -1861,6 +1861,7 @@ static struct
struct redist_region *redist_regs;
u32 nr_redist_regions;
bool single_redist;
+ int enabled_rdists;
u32 maint_irq;
int maint_irq_mode;
phys_addr_t vcpu_base;
@@ -1955,8 +1956,10 @@ static int __init gic_acpi_match_gicc(union acpi_subtable_headers *header,
* If GICC is enabled and has valid gicr base address, then it means
* GICR base is presented via GICC
*/
- if ((gicc->flags & ACPI_MADT_ENABLED) && gicc->gicr_base_address)
+ if ((gicc->flags & ACPI_MADT_ENABLED) && gicc->gicr_base_address) {
+ acpi_data.enabled_rdists++;
return 0;
+ }

/*
* It's perfectly valid firmware can pass disabled GICC entry, driver
@@ -1986,8 +1989,10 @@ static int __init gic_acpi_count_gicr_regions(void)

count = acpi_table_parse_madt(ACPI_MADT_TYPE_GENERIC_INTERRUPT,
gic_acpi_match_gicc, 0);
- if (count > 0)
+ if (count > 0) {
acpi_data.single_redist = true;
+ count = acpi_data.enabled_rdists;
+ }

return count;
}

--
Jazz is not dead. It just smells funny...