iommu/amd: Fix double free of mem-region in error-path
When ioremap_nocache fails in iommu initialization the code calls release_mem_region immediatly. But the function is called again when the propagates into the upper init functions leading to a double-free. Fix that. Reported-by: Don Dutile <ddutile@redhat.com> Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
This commit is contained in:
Родитель
8704a1ba47
Коммит
6e930045ab
|
@ -358,8 +358,6 @@ static void iommu_disable(struct amd_iommu *iommu)
|
|||
*/
|
||||
static u8 * __init iommu_map_mmio_space(u64 address)
|
||||
{
|
||||
u8 *ret;
|
||||
|
||||
if (!request_mem_region(address, MMIO_REGION_LENGTH, "amd_iommu")) {
|
||||
pr_err("AMD-Vi: Can not reserve memory region %llx for mmio\n",
|
||||
address);
|
||||
|
@ -367,13 +365,7 @@ static u8 * __init iommu_map_mmio_space(u64 address)
|
|||
return NULL;
|
||||
}
|
||||
|
||||
ret = ioremap_nocache(address, MMIO_REGION_LENGTH);
|
||||
if (ret != NULL)
|
||||
return ret;
|
||||
|
||||
release_mem_region(address, MMIO_REGION_LENGTH);
|
||||
|
||||
return NULL;
|
||||
return ioremap_nocache(address, MMIO_REGION_LENGTH);
|
||||
}
|
||||
|
||||
static void __init iommu_unmap_mmio_space(struct amd_iommu *iommu)
|
||||
|
|
Загрузка…
Ссылка в новой задаче