mm/hmm: fix memremap.h, move dev_page_fault_t callback to hmm
The kbuild robot reported the following on a development branch that used
memremap.h in a new path:
In file included from arch/m68k/include/asm/pgtable_mm.h:148:0,
from arch/m68k/include/asm/pgtable.h:5,
from include/linux/memremap.h:7,
from drivers//dax/bus.c:3:
arch/m68k/include/asm/motorola_pgtable.h: In function 'pgd_offset':
>> arch/m68k/include/asm/motorola_pgtable.h:199:11: error: dereferencing pointer to incomplete type 'const struct mm_struct'
return mm->pgd + pgd_index(address);
^~
The ->page_fault() callback is specific to HMM. Move it to 'struct
hmm_devmem' where the unusual asm/pgtable.h dependency can be contained in
include/linux/hmm.h. Longer term refactoring this dependency out of HMM
is recommended, but in the meantime memremap.h remains generic.
Link: http://lkml.kernel.org/r/154534090899.3120190.6652620807617715272.stgit@dwillia2-desk3.amr.corp.intel.com
Fixes: 5042db43cc26 ("mm/ZONE_DEVICE: new type of ZONE_DEVICE memory...")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
diff --git a/include/linux/hmm.h b/include/linux/hmm.h
index ed89fbc..66f9ebb 100644
--- a/include/linux/hmm.h
+++ b/include/linux/hmm.h
@@ -69,6 +69,7 @@
#define LINUX_HMM_H
#include <linux/kconfig.h>
+#include <asm/pgtable.h>
#if IS_ENABLED(CONFIG_HMM)
@@ -486,6 +487,7 @@ struct hmm_devmem_ops {
* @device: device to bind resource to
* @ops: memory operations callback
* @ref: per CPU refcount
+ * @page_fault: callback when CPU fault on an unaddressable device page
*
* This an helper structure for device drivers that do not wish to implement
* the gory details related to hotplugging new memoy and allocating struct
@@ -493,7 +495,28 @@ struct hmm_devmem_ops {
*
* Device drivers can directly use ZONE_DEVICE memory on their own if they
* wish to do so.
+ *
+ * The page_fault() callback must migrate page back, from device memory to
+ * system memory, so that the CPU can access it. This might fail for various
+ * reasons (device issues, device have been unplugged, ...). When such error
+ * conditions happen, the page_fault() callback must return VM_FAULT_SIGBUS and
+ * set the CPU page table entry to "poisoned".
+ *
+ * Note that because memory cgroup charges are transferred to the device memory,
+ * this should never fail due to memory restrictions. However, allocation
+ * of a regular system page might still fail because we are out of memory. If
+ * that happens, the page_fault() callback must return VM_FAULT_OOM.
+ *
+ * The page_fault() callback can also try to migrate back multiple pages in one
+ * chunk, as an optimization. It must, however, prioritize the faulting address
+ * over all the others.
*/
+typedef int (*dev_page_fault_t)(struct vm_area_struct *vma,
+ unsigned long addr,
+ const struct page *page,
+ unsigned int flags,
+ pmd_t *pmdp);
+
struct hmm_devmem {
struct completion completion;
unsigned long pfn_first;
@@ -503,6 +526,7 @@ struct hmm_devmem {
struct dev_pagemap pagemap;
const struct hmm_devmem_ops *ops;
struct percpu_ref ref;
+ dev_page_fault_t page_fault;
};
/*