Current code in acpi_os_map_memory uses the direct map in order to map memory
in the low 1MB, but acpi_os_unmap_memory doesn't takes that into account, and
always tries to perform a vunmap, which results in the following WARN:
(XEN) Xen WARN at vmap.c:185
(XEN) [ Xen-4.9-unstable x86_64 debug=y Tainted: C ]
(XEN) CPU:0
(XEN) RIP:e008:[] vmap.c#vm_free+0xd7/0xe0
[...]
(XEN) Xen call trace:
(XEN)[] vmap.c#vm_free+0xd7/0xe0
(XEN)[] acpi_find_root_pointer+0x3a/0x170
(XEN)[] acpi_os_get_root_pointer+0x4e/0x60
(XEN)[] domain_build.c#pvh_setup_acpi_xsdt+0x90/0x240
(XEN)[] domain_build.c#pvh_setup_acpi+0x18a/0x2e0
(XEN)[] domain_build.c#construct_dom0_pvh+0xd2/0x120
(XEN)[] __start_xen+0x1d14/0x2420
(XEN)[] __high_start+0x53/0x60
Fix this by checking if the virtual address passed to acpi_os_unmap_memory
belongs to the direct map.
Signed-off-by: Roger Pau Monné
---
Cc: Jan Beulich
---
xen/drivers/acpi/osl.c | 7 +++
1 file changed, 7 insertions(+)
diff --git a/xen/drivers/acpi/osl.c b/xen/drivers/acpi/osl.c
index 7199047..930e2d9 100644
--- a/xen/drivers/acpi/osl.c
+++ b/xen/drivers/acpi/osl.c
@@ -104,6 +104,13 @@ acpi_os_map_memory(acpi_physical_address phys, acpi_size
size)
void acpi_os_unmap_memory(void __iomem * virt, acpi_size size)
{
+ if (IS_ENABLED(CONFIG_X86) &&
+ (unsigned long)virt >= DIRECTMAP_VIRT_START &&
+ (unsigned long)virt < DIRECTMAP_VIRT_END) {
+ ASSERT(!((__pa(virt) + size - 1) >> 20));
+ return;
+ }
+
if (system_state >= SYS_STATE_boot)
vunmap((void *)((unsigned long)virt & PAGE_MASK));
}
--
2.10.1 (Apple Git-78)
___
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel