Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.kernel > #1322488 > unrolled thread
| Started by | Dan Williams <dan.j.williams@intel.com> |
|---|---|
| First post | 2016-01-30 19:50 +0100 |
| Last post | 2016-01-30 19:50 +0100 |
| Articles | 4 — 1 participant |
Back to article view | Back to linux.kernel
[PATCH 0/3] libnvdimm: fix memmap in pmem support Dan Williams <dan.j.williams@intel.com> - 2016-01-30 19:50 +0100
[PATCH 2/3] libnvdimm, pfn: fix restoring memmap location Dan Williams <dan.j.williams@intel.com> - 2016-01-30 19:50 +0100
[PATCH 1/3] libnvdimm: fix mode determination for e820 devices Dan Williams <dan.j.williams@intel.com> - 2016-01-30 19:50 +0100
[PATCH 3/3] devm_memremap_pages: fix vmem_altmap lifetime + alignment handling Dan Williams <dan.j.williams@intel.com> - 2016-01-30 19:50 +0100
| From | Dan Williams <dan.j.williams@intel.com> |
|---|---|
| Date | 2016-01-30 19:50 +0100 |
| Subject | [PATCH 0/3] libnvdimm: fix memmap in pmem support |
| Message-ID | <qWIKR-81s-3@gated-at.bofh.it> |
Thanks to Jeff for pointing out a thinko in the ndctl implementation of
create-namespace. That bug was covering up the fact that vmem_altmap
handling broke when 'pgmap_radix' was introduced to replace the list
based method [1].
The other two fixes fell out from further expanding the unit tests.
[1]: https://lists.01.org/pipermail/linux-nvdimm/2015-December/003608.html
---
Dan Williams (3):
libnvdimm: fix mode determination for e820 devices
libnvdimm, pfn: fix restoring memmap location
devm_memremap_pages: fix vmem_altmap lifetime + alignment handling
drivers/nvdimm/namespace_devs.c | 8 +++++---
drivers/nvdimm/pfn_devs.c | 4 +---
kernel/memremap.c | 18 +++++++++++-------
3 files changed, 17 insertions(+), 13 deletions(-)
[toc] | [next] | [standalone]
| From | Dan Williams <dan.j.williams@intel.com> |
|---|---|
| Date | 2016-01-30 19:50 +0100 |
| Subject | [PATCH 2/3] libnvdimm, pfn: fix restoring memmap location |
| Message-ID | <qWIKS-81s-15@gated-at.bofh.it> |
| In reply to | #1322488 |
This path was missed when turning on the memmap in pmem support. Permit
'pmem' as a valid location for the map.
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
drivers/nvdimm/pfn_devs.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/drivers/nvdimm/pfn_devs.c b/drivers/nvdimm/pfn_devs.c
index 0cc9048b86e2..ae81a2f1da50 100644
--- a/drivers/nvdimm/pfn_devs.c
+++ b/drivers/nvdimm/pfn_devs.c
@@ -301,10 +301,8 @@ int nd_pfn_validate(struct nd_pfn *nd_pfn)
switch (le32_to_cpu(pfn_sb->mode)) {
case PFN_MODE_RAM:
- break;
case PFN_MODE_PMEM:
- /* TODO: allocate from PMEM support */
- return -ENOTTY;
+ break;
default:
return -ENXIO;
}
[toc] | [prev] | [next] | [standalone]
| From | Dan Williams <dan.j.williams@intel.com> |
|---|---|
| Date | 2016-01-30 19:50 +0100 |
| Subject | [PATCH 1/3] libnvdimm: fix mode determination for e820 devices |
| Message-ID | <qWIKR-81s-7@gated-at.bofh.it> |
| In reply to | #1322488 |
Correctly display "safe" mode when a btt is established on a e820/memmap defined pmem namespace. Signed-off-by: Dan Williams <dan.j.williams@intel.com> --- drivers/nvdimm/namespace_devs.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/drivers/nvdimm/namespace_devs.c b/drivers/nvdimm/namespace_devs.c index 8ebfcaae3f5a..9edf7eb7d17c 100644 --- a/drivers/nvdimm/namespace_devs.c +++ b/drivers/nvdimm/namespace_devs.c @@ -1277,10 +1277,12 @@ static ssize_t mode_show(struct device *dev, device_lock(dev); claim = ndns->claim; - if (pmem_should_map_pages(dev) || (claim && is_nd_pfn(claim))) - mode = "memory"; - else if (claim && is_nd_btt(claim)) + if (claim && is_nd_btt(claim)) mode = "safe"; + else if (claim && is_nd_pfn(claim)) + mode = "memory"; + else if (!claim && pmem_should_map_pages(dev)) + mode = "memory"; else mode = "raw"; rc = sprintf(buf, "%s\n", mode);
[toc] | [prev] | [next] | [standalone]
| From | Dan Williams <dan.j.williams@intel.com> |
|---|---|
| Date | 2016-01-30 19:50 +0100 |
| Subject | [PATCH 3/3] devm_memremap_pages: fix vmem_altmap lifetime + alignment handling |
| Message-ID | <qWIKS-81s-17@gated-at.bofh.it> |
| In reply to | #1322488 |
to_vmem_altmap() needs to return valid results until
arch_remove_memory() completes. It also needs to be valid for any pfn
in a section regardless of whether that pfn maps to data. This escape
was a result of a bug in the unit test.
The signature of this bug is that free_pagetable() fails to retrieve a
vmem_altmap and goes off into the weeds:
BUG: unable to handle kernel NULL pointer dereference at (null)
IP: [<ffffffff811d2629>] get_pfnblock_flags_mask+0x49/0x60
[..]
Call Trace:
[<ffffffff811d3477>] free_hot_cold_page+0x97/0x1d0
[<ffffffff811d367a>] __free_pages+0x2a/0x40
[<ffffffff8191e669>] free_pagetable+0x8c/0xd4
[<ffffffff8191ef4e>] remove_pagetable+0x37a/0x808
[<ffffffff8191b210>] vmemmap_free+0x10/0x20
Fixes: 4b94ffdc4163 ("x86, mm: introduce vmem_altmap to augment vmemmap_populate()")
Cc: Andrew Morton <akpm@linux-foundation.org>
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
kernel/memremap.c | 18 +++++++++++-------
1 file changed, 11 insertions(+), 7 deletions(-)
diff --git a/kernel/memremap.c b/kernel/memremap.c
index e517a16cb426..cbc3e97e2bb4 100644
--- a/kernel/memremap.c
+++ b/kernel/memremap.c
@@ -183,7 +183,11 @@ EXPORT_SYMBOL(put_zone_device_page);
static void pgmap_radix_release(struct resource *res)
{
- resource_size_t key;
+ resource_size_t key, align_start, align_size, align_end;
+
+ align_start = res->start & ~(SECTION_SIZE - 1);
+ align_size = ALIGN(resource_size(res), SECTION_SIZE);
+ align_end = align_start + align_size - 1;
mutex_lock(&pgmap_lock);
for (key = res->start; key <= res->end; key += SECTION_SIZE)
@@ -226,12 +230,11 @@ static void devm_memremap_pages_release(struct device *dev, void *data)
percpu_ref_put(pgmap->ref);
}
- pgmap_radix_release(res);
-
/* pages are dead and unused, undo the arch mapping */
align_start = res->start & ~(SECTION_SIZE - 1);
align_size = ALIGN(resource_size(res), SECTION_SIZE);
arch_remove_memory(align_start, align_size);
+ pgmap_radix_release(res);
dev_WARN_ONCE(dev, pgmap->altmap && pgmap->altmap->alloc,
"%s: failed to free all reserved pages\n", __func__);
}
@@ -267,7 +270,7 @@ void *devm_memremap_pages(struct device *dev, struct resource *res,
{
int is_ram = region_intersects(res->start, resource_size(res),
"System RAM");
- resource_size_t key, align_start, align_size;
+ resource_size_t key, align_start, align_size, align_end;
struct dev_pagemap *pgmap;
struct page_map *page_map;
unsigned long pfn;
@@ -309,7 +312,10 @@ void *devm_memremap_pages(struct device *dev, struct resource *res,
mutex_lock(&pgmap_lock);
error = 0;
- for (key = res->start; key <= res->end; key += SECTION_SIZE) {
+ align_start = res->start & ~(SECTION_SIZE - 1);
+ align_size = ALIGN(resource_size(res), SECTION_SIZE);
+ align_end = align_start + align_size - 1;
+ for (key = align_start; key <= align_end; key += SECTION_SIZE) {
struct dev_pagemap *dup;
rcu_read_lock();
@@ -336,8 +342,6 @@ void *devm_memremap_pages(struct device *dev, struct resource *res,
if (nid < 0)
nid = numa_mem_id();
- align_start = res->start & ~(SECTION_SIZE - 1);
- align_size = ALIGN(resource_size(res), SECTION_SIZE);
error = arch_add_memory(nid, align_start, align_size, true);
if (error)
goto err_add_memory;
[toc] | [prev] | [standalone]
Back to top | Article view | linux.kernel
csiph-web