Re: [HMM 07/15] mm/ZONE_DEVICE: new type of ZONE_DEVICE for unaddressable memory v3
From: John Hubbard
Date: Thu Jun 08 2017 - 23:55:15 EST
On 05/24/2017 10:20 AM, JÃrÃme Glisse wrote:
[...8<...]
+#if IS_ENABLED(CONFIG_DEVICE_PRIVATE)
+int device_private_entry_fault(struct vm_area_struct *vma,
+ unsigned long addr,
+ swp_entry_t entry,
+ unsigned int flags,
+ pmd_t *pmdp)
+{
+ struct page *page = device_private_entry_to_page(entry);
+
+ /*
+ * The page_fault() callback must migrate page back to system memory
+ * so that CPU can access it. This might fail for various reasons
+ * (device issue, device was unsafely unplugged, ...). When such
+ * error conditions happen, the callback must return VM_FAULT_SIGBUS.
+ *
+ * Note that because memory cgroup charges are accounted to the device
+ * memory, this should never fail because of memory restrictions (but
+ * allocation of regular system page might still fail because we are
+ * out of memory).
+ *
+ * There is a more in-depth description of what that callback can and
+ * cannot do, in include/linux/memremap.h
+ */
+ return page->pgmap->page_fault(vma, addr, page, flags, pmdp);
+}
+EXPORT_SYMBOL(device_private_entry_fault);
+#endif /* CONFIG_DEVICE_PRIVATE */
+
static void pgmap_radix_release(struct resource *res)
{
resource_size_t key, align_start, align_size, align_end;
@@ -321,6 +351,10 @@ void *devm_memremap_pages(struct device *dev, struct resource *res,
}
pgmap->ref = ref;
pgmap->res = &page_map->res;
+ pgmap->type = MEMORY_DEVICE_PUBLIC;
+ pgmap->page_fault = NULL;
+ pgmap->page_free = NULL;
+ pgmap->data = NULL;
mutex_lock(&pgmap_lock);
error = 0;
diff --git a/mm/Kconfig b/mm/Kconfig
index d744cff..f5357ff 100644
--- a/mm/Kconfig
+++ b/mm/Kconfig
@@ -736,6 +736,19 @@ config ZONE_DEVICE
If FS_DAX is enabled, then say Y.
+config DEVICE_PRIVATE
+ bool "Unaddressable device memory (GPU memory, ...)"
+ depends on X86_64
+ depends on ZONE_DEVICE
+ depends on MEMORY_HOTPLUG
+ depends on MEMORY_HOTREMOVE
+ depends on SPARSEMEM_VMEMMAP
+
+ help
+ Allows creation of struct pages to represent unaddressable device
+ memory; i.e., memory that is only accessible from the device (or
+ group of devices).
+
Hi Jerome,
CONFIG_DEVICE_PRIVATE has caused me some problems, because it's not coupled to HMM_DEVMEM.
To fix this, my first choice would be to just s/DEVICE_PRIVATE/HMM_DEVMEM/g , because I don't see
any value to DEVICE_PRIVATE as an independent Kconfig choice. It's complicating the Kconfig choices,
and adding problems. However, if DEVICE_PRIVATE must be kept, then something like this also fixes my
HMM tests:
From: John Hubbard <jhubbard@xxxxxxxxxx>
Date: Thu, 8 Jun 2017 20:13:13 -0700
Subject: [PATCH] hmm: select CONFIG_DEVICE_PRIVATE with HMM_DEVMEM
The HMM_DEVMEM feature is useless without the various
features that are guarded with CONFIG_DEVICE_PRIVATE.
Therefore, auto-select DEVICE_PRIVATE when selecting
HMM_DEVMEM.
Otherwise, you can easily end up with a partially
working HMM installation: if you select HMM_DEVMEM,
but do not select DEVICE_PRIVATE, then faulting and
migrating to a device (such as a GPU) works, but CPU
page faults are ignored, so the page never migrates
back to the CPU.
Signed-off-by: John Hubbard <jhubbard@xxxxxxxxxx>
---
mm/Kconfig | 2 ++
1 file changed, 2 insertions(+)
diff --git a/mm/Kconfig b/mm/Kconfig
index 46296d5d7570..23d2f5ec865e 100644
--- a/mm/Kconfig
+++ b/mm/Kconfig
@@ -318,6 +318,8 @@ config HMM_DEVMEM
bool "HMM device memory helpers (to leverage ZONE_DEVICE)"
depends on ARCH_HAS_HMM
select HMM
+ select DEVICE_PRIVATE
+
help
HMM devmem is a set of helper routines to leverage the ZONE_DEVICE
feature. This is just to avoid having device drivers to replicating a lot
--
2.13.1
This is a minor thing, and I don't think this needs to hold up merging HMM v23 into -mm, IMHO. But I
would like it fixed at some point.
thanks,
--
John Hubbard
NVIDIA