ACPI/IORT: Check node revision for PMCG resources
authorRobin Murphy <robin.murphy@arm.com>
Thu, 3 Feb 2022 19:31:24 +0000 (19:31 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 16 Feb 2022 11:56:19 +0000 (12:56 +0100)
commit da5fb9e1ad3fbf632dce735f1bdad257ca528499 upstream.

The original version of the IORT PMCG definition had an oversight
wherein there was no way to describe the second register page for an
implementation using the recommended RELOC_CTRS feature. Although the
spec was fixed, and the final patches merged to ACPICA and Linux written
against the new version, it seems that some old firmware based on the
original revision has survived and turned up in the wild.

Add a check for the original PMCG definition, and avoid filling in the
second memory resource with nonsense if so. Otherwise it is likely that
something horrible will happen when the PMCG driver attempts to probe.

Reported-by: Michael Petlan <mpetlan@redhat.com>
Fixes: 24e516049360 ("ACPI/IORT: Add support for PMCG")
Cc: <stable@vger.kernel.org> # 5.2.x
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Acked-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Link: https://lore.kernel.org/r/75628ae41c257fb73588f7bf1c4459160e04be2b.1643916258.git.robin.murphy@arm.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/acpi/arm64/iort.c

index 3b23fb7..f2f8f05 100644 (file)
@@ -1361,9 +1361,17 @@ static void __init arm_smmu_v3_pmcg_init_resources(struct resource *res,
        res[0].start = pmcg->page0_base_address;
        res[0].end = pmcg->page0_base_address + SZ_4K - 1;
        res[0].flags = IORESOURCE_MEM;
-       res[1].start = pmcg->page1_base_address;
-       res[1].end = pmcg->page1_base_address + SZ_4K - 1;
-       res[1].flags = IORESOURCE_MEM;
+       /*
+        * The initial version in DEN0049C lacked a way to describe register
+        * page 1, which makes it broken for most PMCG implementations; in
+        * that case, just let the driver fail gracefully if it expects to
+        * find a second memory resource.
+        */
+       if (node->revision > 0) {
+               res[1].start = pmcg->page1_base_address;
+               res[1].end = pmcg->page1_base_address + SZ_4K - 1;
+               res[1].flags = IORESOURCE_MEM;
+       }
 
        if (pmcg->overflow_gsiv)
                acpi_iort_register_irq(pmcg->overflow_gsiv, "overflow",