mm/memory_hotplug: add pgprot_t to mhp_params

devm_memremap_pages() is currently used by the PCI P2PDMA code to create struct page mappings for IO memory. At present, these mappings are created with PAGE_KERNEL which implies setting the PAT bits to be WB. However, on x86, an mtrr register will typically override this and force the cache type to be UC-. In the case firmware doesn't set this register it is effectively WB and will typically result in a machine check exception when it's accessed. Other arches are not currently likely to function correctly seeing they don't have any MTRR registers to fall back on. To solve this, provide a way to specify the pgprot value explicitly to arch_add_memory(). Of the arches that support MEMORY_HOTPLUG: x86_64, and arm64 need a simple change to pass the pgprot_t down to their respective functions which set up the page tables. For x86_32, set the page tables explicitly using _set_memory_prot() (seeing they are already mapped). For ia64, s390 and sh, reject anything but PAGE_KERNEL settings -- this should be fine, for now, seeing these architectures don't support ZONE_DEVICE. A check in __add_pages() is also added to ensure the pgprot parameter was set for all arches.

Logan Gunthorpe committed 6 years ago

Commit bfeb0221 parent 4e00c5a10 files · +36 −7Browse files

Unified · Split

10 files

arch/arm64/mm/mmu.c +2 −1
arch/ia64/mm/init.c +3 −0
arch/powerpc/mm/mem.c +2 −1
arch/s390/mm/init.c +3 −0
arch/sh/mm/init.c +3 −0
arch/x86/mm/init_32.c +12 −0
arch/x86/mm/init_64.c +1 −1
include/linux/memory_hotplug.h +3 −0
mm/memory_hotplug.c +4 −1
mm/memremap.c +3 −3

arch/arm64/mm/mmu.c+2−1

@@ -1382,7 +1382,8 @@ int arch_add_memory(int nid, u64 start, u64 size, · 1381 unmodified lines
1382	1382	flags = NO_BLOCK_MAPPINGS \| NO_CONT_MAPPINGS;
1383	1383
1384	1384	__create_pgd_mapping(swapper_pg_dir, start, __phys_to_virt(start),
1385		− size, PAGE_KERNEL, __pgd_pgtable_alloc, flags);
	1385	+ size, params->pgprot, __pgd_pgtable_alloc,
	1386	+ flags);
1386	1387
1387	1388	memblock_clear_nomap(start, size);
1388	1389

arch/ia64/mm/init.c+3−0

@@ -676,6 +676,9 @@ int arch_add_memory(int nid, u64 start, u64 size, · 675 unmodified lines
676	676	unsigned long nr_pages = size >> PAGE_SHIFT;
677	677	int ret;
678	678
	679	+ if (WARN_ON_ONCE(params->pgprot.pgprot != PAGE_KERNEL.pgprot))
	680	+ return -EINVAL;
	681	+
679	682	ret = __add_pages(nid, start_pfn, nr_pages, params);
680	683	if (ret)
681	684	printk("%s: Problem encountered in __add_pages() as ret=%d\n",

arch/powerpc/mm/mem.c+2−1

@@ -132,7 +132,8 @@ int __ref arch_add_memory(int nid, u64 start, u64 size, · 131 unmodified lines
132	132	resize_hpt_for_hotplug(memblock_phys_mem_size());
133	133
134	134	start = (unsigned long)__va(start);
135		− rc = create_section_mapping(start, start + size, nid, PAGE_KERNEL);
	135	+ rc = create_section_mapping(start, start + size, nid,
	136	+ params->pgprot);
136	137	if (rc) {
137	138	pr_warn("Unable to create mapping for hot added memory 0x%llx..0x%llx: %d\n",
138	139	start, start + size, rc);

arch/s390/mm/init.c+3−0

@@ -277,6 +277,9 @@ int arch_add_memory(int nid, u64 start, u64 size, · 276 unmodified lines
277	277	if (WARN_ON_ONCE(params->altmap))
278	278	return -EINVAL;
279	279
	280	+ if (WARN_ON_ONCE(params->pgprot.pgprot != PAGE_KERNEL.pgprot))
	281	+ return -EINVAL;
	282	+
280	283	rc = vmem_add_mapping(start, size);
281	284	if (rc)
282	285	return rc;

arch/sh/mm/init.c+3−0

@@ -412,6 +412,9 @@ int arch_add_memory(int nid, u64 start, u64 size, · 411 unmodified lines
412	412	unsigned long nr_pages = size >> PAGE_SHIFT;
413	413	int ret;
414	414
	415	+ if (WARN_ON_ONCE(params->pgprot.pgprot != PAGE_KERNEL.pgprot)
	416	+ return -EINVAL;
	417	+
415	418	/* We only have ZONE_NORMAL, so this is easy.. */
416	419	ret = __add_pages(nid, start_pfn, nr_pages, params);
417	420	if (unlikely(ret))

arch/x86/mm/init_32.c+12−0

@@ -824,6 +824,18 @@ int arch_add_memory(int nid, u64 start, u64 size, · 823 unmodified lines
824	824	{
825	825	unsigned long start_pfn = start >> PAGE_SHIFT;
826	826	unsigned long nr_pages = size >> PAGE_SHIFT;
	827	+ int ret;
	828	+
	829	+ /*
	830	+ * The page tables were already mapped at boot so if the caller
	831	+ * requests a different mapping type then we must change all the
	832	+ * pages with __set_memory_prot().
	833	+ */
	834	+ if (params->pgprot.pgprot != PAGE_KERNEL.pgprot) {
	835	+ ret = __set_memory_prot(start, nr_pages, params->pgprot);
	836	+ if (ret)
	837	+ return ret;
	838	+ }
827	839
828	840	return __add_pages(nid, start_pfn, nr_pages, params);
829	841	}

arch/x86/mm/init_64.c+1−1

@@ -867,7 +867,7 @@ int arch_add_memory(int nid, u64 start, u64 size, · 866 unmodified lines
867	867	unsigned long start_pfn = start >> PAGE_SHIFT;
868	868	unsigned long nr_pages = size >> PAGE_SHIFT;
869	869
870		− init_memory_mapping(start, start + size, PAGE_KERNEL);
	870	+ init_memory_mapping(start, start + size, params->pgprot);
871	871
872	872	return add_pages(nid, start_pfn, nr_pages, params);
873	873	}

include/linux/memory_hotplug.h+3−0

@@ -60,9 +60,12 @@ enum { · 59 unmodified lines
60	60	/*
61	61	* Extended parameters for memory hotplug:
62	62	* altmap: alternative allocator for memmap array (optional)
	63	+ * pgprot: page protection flags to apply to newly created page tables
	64	+ * (required)
63	65	*/
64	66	struct mhp_params {
65	67	struct vmem_altmap *altmap;
	68	+ pgprot_t pgprot;
66	69	};
67	70
68	71	/*

mm/memory_hotplug.c+4−1

@@ -311,6 +311,9 @@ int __ref __add_pages(int nid, unsigned long pfn, unsigned long nr_pages, · 310 unmodified lines
311	311	int err;
312	312	struct vmem_altmap *altmap = params->altmap;
313	313
	314	+ if (WARN_ON_ONCE(!params->pgprot.pgprot))
	315	+ return -EINVAL;
	316	+
314	317	err = check_hotplug_memory_addressable(pfn, nr_pages);
315	318	if (err)
316	319	return err;
@@ -1002,7 +1005,7 @@ static int online_memory_block(struct memory_block mem, void arg) · 685 unmodified lines
1002	1005	*/
1003	1006	int __ref add_memory_resource(int nid, struct resource *res)
1004	1007	{
1005		− struct mhp_params params = {};
	1008	+ struct mhp_params params = { .pgprot = PAGE_KERNEL };
1006	1009	u64 start, size;
1007	1010	bool new_node = false;
1008	1011	int ret;

mm/memremap.c+3−3

@@ -189,8 +189,8 @@ void memremap_pages(struct dev_pagemap pgmap, int nid) · 188 unmodified lines
189	189	* We do not want any optional features only our own memmap
190	190	*/
191	191	.altmap = pgmap_altmap(pgmap),
	192	+ .pgprot = PAGE_KERNEL,
192	193	};
193		− pgprot_t pgprot = PAGE_KERNEL;
194	194	int error, is_ram;
195	195	bool need_devmap_managed = true;
196	196
@@ -282,8 +282,8 @@ void memremap_pages(struct dev_pagemap pgmap, int nid) · 85 unmodified lines
282	282	if (nid < 0)
283	283	nid = numa_mem_id();
284	284
285		− error = track_pfn_remap(NULL, &pgprot, PHYS_PFN(res->start), 0,
286		− resource_size(res));
	285	+ error = track_pfn_remap(NULL, &params.pgprot, PHYS_PFN(res->start),
	286	+ 0, resource_size(res));
287	287	if (error)
288	288	goto err_pfn_remap;
289	289