2019-05-29 14:18:02 +00:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-only
|
2015-04-01 07:12:19 +00:00
|
|
|
/*
|
|
|
|
* Persistent Memory Driver
|
|
|
|
*
|
2015-06-09 19:33:45 +00:00
|
|
|
* Copyright (c) 2014-2015, Intel Corporation.
|
2015-04-01 07:12:19 +00:00
|
|
|
* Copyright (c) 2015, Christoph Hellwig <hch@lst.de>.
|
|
|
|
* Copyright (c) 2015, Boaz Harrosh <boaz@plexistor.com>.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <linux/hdreg.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/platform_device.h>
|
2018-07-14 04:50:37 +00:00
|
|
|
#include <linux/set_memory.h>
|
2015-04-01 07:12:19 +00:00
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/moduleparam.h>
|
2016-01-05 07:50:23 +00:00
|
|
|
#include <linux/badblocks.h>
|
2016-01-16 00:56:19 +00:00
|
|
|
#include <linux/memremap.h>
|
2015-08-01 06:16:37 +00:00
|
|
|
#include <linux/vmalloc.h>
|
2017-04-28 17:23:37 +00:00
|
|
|
#include <linux/blk-mq.h>
|
2016-01-16 00:56:14 +00:00
|
|
|
#include <linux/pfn_t.h>
|
2015-04-01 07:12:19 +00:00
|
|
|
#include <linux/slab.h>
|
2017-05-29 19:22:50 +00:00
|
|
|
#include <linux/uio.h>
|
2017-01-25 07:02:09 +00:00
|
|
|
#include <linux/dax.h>
|
2015-06-09 19:33:45 +00:00
|
|
|
#include <linux/nd.h>
|
2017-11-16 01:33:00 +00:00
|
|
|
#include <linux/backing-dev.h>
|
2020-06-08 04:41:42 +00:00
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <asm/cacheflush.h>
|
2016-06-17 18:08:06 +00:00
|
|
|
#include "pmem.h"
|
2015-08-01 06:16:37 +00:00
|
|
|
#include "pfn.h"
|
2015-06-09 19:33:45 +00:00
|
|
|
#include "nd.h"
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2016-07-08 02:44:50 +00:00
|
|
|
static struct device *to_dev(struct pmem_device *pmem)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* nvdimm bus services need a 'dev' parameter, and we record the device
|
|
|
|
* at init in bb.dev.
|
|
|
|
*/
|
|
|
|
return pmem->bb.dev;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct nd_region *to_region(struct pmem_device *pmem)
|
|
|
|
{
|
|
|
|
return to_nd_region(to_dev(pmem)->parent);
|
|
|
|
}
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2018-07-14 04:50:37 +00:00
|
|
|
static void hwpoison_clear(struct pmem_device *pmem,
|
|
|
|
phys_addr_t phys, unsigned int len)
|
|
|
|
{
|
|
|
|
unsigned long pfn_start, pfn_end, pfn;
|
|
|
|
|
|
|
|
/* only pmem in the linear map supports HWPoison */
|
|
|
|
if (is_vmalloc_addr(pmem->virt_addr))
|
|
|
|
return;
|
|
|
|
|
|
|
|
pfn_start = PHYS_PFN(phys);
|
|
|
|
pfn_end = pfn_start + PHYS_PFN(len);
|
|
|
|
for (pfn = pfn_start; pfn < pfn_end; pfn++) {
|
|
|
|
struct page *page = pfn_to_page(pfn);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Note, no need to hold a get_dev_pagemap() reference
|
|
|
|
* here since we're in the driver I/O path and
|
|
|
|
* outstanding I/O requests pin the dev_pagemap.
|
|
|
|
*/
|
|
|
|
if (test_and_clear_pmem_poison(page))
|
|
|
|
clear_mce_nospec(pfn);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2017-06-03 07:38:06 +00:00
|
|
|
static blk_status_t pmem_clear_poison(struct pmem_device *pmem,
|
|
|
|
phys_addr_t offset, unsigned int len)
|
2016-03-08 15:16:07 +00:00
|
|
|
{
|
2016-07-08 02:44:50 +00:00
|
|
|
struct device *dev = to_dev(pmem);
|
2016-03-08 15:16:07 +00:00
|
|
|
sector_t sector;
|
|
|
|
long cleared;
|
2017-06-03 07:38:06 +00:00
|
|
|
blk_status_t rc = BLK_STS_OK;
|
2016-03-08 15:16:07 +00:00
|
|
|
|
|
|
|
sector = (offset - pmem->data_offset) / 512;
|
|
|
|
|
2016-12-16 16:10:31 +00:00
|
|
|
cleared = nvdimm_clear_poison(dev, pmem->phys_addr + offset, len);
|
|
|
|
if (cleared < len)
|
2017-06-03 07:38:06 +00:00
|
|
|
rc = BLK_STS_IOERR;
|
2016-03-08 15:16:07 +00:00
|
|
|
if (cleared > 0 && cleared / 512) {
|
2018-07-14 04:50:37 +00:00
|
|
|
hwpoison_clear(pmem, pmem->phys_addr + offset, cleared);
|
2016-12-16 16:10:31 +00:00
|
|
|
cleared /= 512;
|
2018-03-06 00:39:31 +00:00
|
|
|
dev_dbg(dev, "%#llx clear %ld sector%s\n",
|
2016-12-16 16:10:31 +00:00
|
|
|
(unsigned long long) sector, cleared,
|
|
|
|
cleared > 1 ? "s" : "");
|
2016-12-04 18:48:58 +00:00
|
|
|
badblocks_clear(&pmem->bb, sector, cleared);
|
2017-06-12 22:25:11 +00:00
|
|
|
if (pmem->bb_state)
|
|
|
|
sysfs_notify_dirent(pmem->bb_state);
|
2016-03-08 15:16:07 +00:00
|
|
|
}
|
2016-10-13 15:54:21 +00:00
|
|
|
|
2017-05-30 06:00:34 +00:00
|
|
|
arch_invalidate_pmem(pmem->virt_addr + offset, len);
|
2016-12-16 16:10:31 +00:00
|
|
|
|
|
|
|
return rc;
|
2016-03-08 15:16:07 +00:00
|
|
|
}
|
|
|
|
|
2016-09-30 23:19:30 +00:00
|
|
|
static void write_pmem(void *pmem_addr, struct page *page,
|
|
|
|
unsigned int off, unsigned int len)
|
|
|
|
{
|
2017-09-06 23:22:27 +00:00
|
|
|
unsigned int chunk;
|
|
|
|
void *mem;
|
|
|
|
|
|
|
|
while (len) {
|
|
|
|
mem = kmap_atomic(page);
|
2019-04-04 02:58:01 +00:00
|
|
|
chunk = min_t(unsigned int, len, PAGE_SIZE - off);
|
2017-09-06 23:22:27 +00:00
|
|
|
memcpy_flushcache(pmem_addr, mem + off, chunk);
|
|
|
|
kunmap_atomic(mem);
|
|
|
|
len -= chunk;
|
|
|
|
off = 0;
|
|
|
|
page++;
|
2019-04-04 02:58:01 +00:00
|
|
|
pmem_addr += chunk;
|
2017-09-06 23:22:27 +00:00
|
|
|
}
|
2016-09-30 23:19:30 +00:00
|
|
|
}
|
|
|
|
|
2017-06-03 07:38:06 +00:00
|
|
|
static blk_status_t read_pmem(struct page *page, unsigned int off,
|
2016-09-30 23:19:30 +00:00
|
|
|
void *pmem_addr, unsigned int len)
|
|
|
|
{
|
2017-09-06 23:22:27 +00:00
|
|
|
unsigned int chunk;
|
2018-05-04 00:06:21 +00:00
|
|
|
unsigned long rem;
|
2017-09-06 23:22:27 +00:00
|
|
|
void *mem;
|
|
|
|
|
|
|
|
while (len) {
|
|
|
|
mem = kmap_atomic(page);
|
2019-04-04 02:58:01 +00:00
|
|
|
chunk = min_t(unsigned int, len, PAGE_SIZE - off);
|
x86, powerpc: Rename memcpy_mcsafe() to copy_mc_to_{user, kernel}()
In reaction to a proposal to introduce a memcpy_mcsafe_fast()
implementation Linus points out that memcpy_mcsafe() is poorly named
relative to communicating the scope of the interface. Specifically what
addresses are valid to pass as source, destination, and what faults /
exceptions are handled.
Of particular concern is that even though x86 might be able to handle
the semantics of copy_mc_to_user() with its common copy_user_generic()
implementation other archs likely need / want an explicit path for this
case:
On Fri, May 1, 2020 at 11:28 AM Linus Torvalds <torvalds@linux-foundation.org> wrote:
>
> On Thu, Apr 30, 2020 at 6:21 PM Dan Williams <dan.j.williams@intel.com> wrote:
> >
> > However now I see that copy_user_generic() works for the wrong reason.
> > It works because the exception on the source address due to poison
> > looks no different than a write fault on the user address to the
> > caller, it's still just a short copy. So it makes copy_to_user() work
> > for the wrong reason relative to the name.
>
> Right.
>
> And it won't work that way on other architectures. On x86, we have a
> generic function that can take faults on either side, and we use it
> for both cases (and for the "in_user" case too), but that's an
> artifact of the architecture oddity.
>
> In fact, it's probably wrong even on x86 - because it can hide bugs -
> but writing those things is painful enough that everybody prefers
> having just one function.
Replace a single top-level memcpy_mcsafe() with either
copy_mc_to_user(), or copy_mc_to_kernel().
Introduce an x86 copy_mc_fragile() name as the rename for the
low-level x86 implementation formerly named memcpy_mcsafe(). It is used
as the slow / careful backend that is supplanted by a fast
copy_mc_generic() in a follow-on patch.
One side-effect of this reorganization is that separating copy_mc_64.S
to its own file means that perf no longer needs to track dependencies
for its memcpy_64.S benchmarks.
[ bp: Massage a bit. ]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Cc: <stable@vger.kernel.org>
Link: http://lore.kernel.org/r/CAHk-=wjSqtXAqfUJxFtWNwmguFASTgB0dz1dT3V-78Quiezqbg@mail.gmail.com
Link: https://lkml.kernel.org/r/160195561680.2163339.11574962055305783722.stgit@dwillia2-desk3.amr.corp.intel.com
2020-10-06 03:40:16 +00:00
|
|
|
rem = copy_mc_to_kernel(mem + off, pmem_addr, chunk);
|
2017-09-06 23:22:27 +00:00
|
|
|
kunmap_atomic(mem);
|
2018-05-04 00:06:21 +00:00
|
|
|
if (rem)
|
2017-09-06 23:22:27 +00:00
|
|
|
return BLK_STS_IOERR;
|
|
|
|
len -= chunk;
|
|
|
|
off = 0;
|
|
|
|
page++;
|
2019-04-04 02:58:01 +00:00
|
|
|
pmem_addr += chunk;
|
2017-09-06 23:22:27 +00:00
|
|
|
}
|
2017-06-03 07:38:06 +00:00
|
|
|
return BLK_STS_OK;
|
2016-09-30 23:19:30 +00:00
|
|
|
}
|
|
|
|
|
2020-02-28 16:34:51 +00:00
|
|
|
static blk_status_t pmem_do_read(struct pmem_device *pmem,
|
|
|
|
struct page *page, unsigned int page_off,
|
|
|
|
sector_t sector, unsigned int len)
|
|
|
|
{
|
|
|
|
blk_status_t rc;
|
|
|
|
phys_addr_t pmem_off = sector * 512 + pmem->data_offset;
|
|
|
|
void *pmem_addr = pmem->virt_addr + pmem_off;
|
|
|
|
|
|
|
|
if (unlikely(is_bad_pmem(&pmem->bb, sector, len)))
|
|
|
|
return BLK_STS_IOERR;
|
|
|
|
|
|
|
|
rc = read_pmem(page, page_off, pmem_addr, len);
|
|
|
|
flush_dcache_page(page);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
static blk_status_t pmem_do_write(struct pmem_device *pmem,
|
|
|
|
struct page *page, unsigned int page_off,
|
|
|
|
sector_t sector, unsigned int len)
|
2015-04-01 07:12:19 +00:00
|
|
|
{
|
2017-06-03 07:38:06 +00:00
|
|
|
blk_status_t rc = BLK_STS_OK;
|
2016-03-08 15:16:07 +00:00
|
|
|
bool bad_pmem = false;
|
2015-08-01 06:16:37 +00:00
|
|
|
phys_addr_t pmem_off = sector * 512 + pmem->data_offset;
|
2016-06-04 01:06:47 +00:00
|
|
|
void *pmem_addr = pmem->virt_addr + pmem_off;
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2016-03-08 15:16:07 +00:00
|
|
|
if (unlikely(is_bad_pmem(&pmem->bb, sector, len)))
|
|
|
|
bad_pmem = true;
|
|
|
|
|
2020-02-28 16:34:51 +00:00
|
|
|
/*
|
|
|
|
* Note that we write the data both before and after
|
|
|
|
* clearing poison. The write before clear poison
|
|
|
|
* handles situations where the latest written data is
|
|
|
|
* preserved and the clear poison operation simply marks
|
|
|
|
* the address range as valid without changing the data.
|
|
|
|
* In this case application software can assume that an
|
|
|
|
* interrupted write will either return the new good
|
|
|
|
* data or an error.
|
|
|
|
*
|
|
|
|
* However, if pmem_clear_poison() leaves the data in an
|
|
|
|
* indeterminate state we need to perform the write
|
|
|
|
* after clear poison.
|
|
|
|
*/
|
|
|
|
flush_dcache_page(page);
|
|
|
|
write_pmem(pmem_addr, page, page_off, len);
|
|
|
|
if (unlikely(bad_pmem)) {
|
|
|
|
rc = pmem_clear_poison(pmem, pmem_off, len);
|
|
|
|
write_pmem(pmem_addr, page, page_off, len);
|
2015-04-01 07:12:19 +00:00
|
|
|
}
|
|
|
|
|
2016-03-06 23:20:51 +00:00
|
|
|
return rc;
|
2015-04-01 07:12:19 +00:00
|
|
|
}
|
|
|
|
|
2020-07-01 08:59:43 +00:00
|
|
|
static blk_qc_t pmem_submit_bio(struct bio *bio)
|
2015-04-01 07:12:19 +00:00
|
|
|
{
|
2019-07-05 14:03:22 +00:00
|
|
|
int ret = 0;
|
2017-06-03 07:38:06 +00:00
|
|
|
blk_status_t rc = 0;
|
2015-05-16 16:28:53 +00:00
|
|
|
bool do_acct;
|
|
|
|
unsigned long start;
|
2015-04-01 07:12:19 +00:00
|
|
|
struct bio_vec bvec;
|
|
|
|
struct bvec_iter iter;
|
2020-05-08 16:15:17 +00:00
|
|
|
struct pmem_device *pmem = bio->bi_disk->private_data;
|
2016-06-02 03:48:15 +00:00
|
|
|
struct nd_region *nd_region = to_region(pmem);
|
|
|
|
|
2018-06-06 16:45:12 +00:00
|
|
|
if (bio->bi_opf & REQ_PREFLUSH)
|
2019-07-05 14:03:22 +00:00
|
|
|
ret = nvdimm_flush(nd_region, bio);
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2020-05-27 05:24:10 +00:00
|
|
|
do_acct = blk_queue_io_stat(bio->bi_disk->queue);
|
|
|
|
if (do_acct)
|
|
|
|
start = bio_start_io_acct(bio);
|
2016-01-06 20:03:41 +00:00
|
|
|
bio_for_each_segment(bvec, bio, iter) {
|
2020-02-28 16:34:51 +00:00
|
|
|
if (op_is_write(bio_op(bio)))
|
|
|
|
rc = pmem_do_write(pmem, bvec.bv_page, bvec.bv_offset,
|
|
|
|
iter.bi_sector, bvec.bv_len);
|
|
|
|
else
|
|
|
|
rc = pmem_do_read(pmem, bvec.bv_page, bvec.bv_offset,
|
|
|
|
iter.bi_sector, bvec.bv_len);
|
2016-01-06 20:03:41 +00:00
|
|
|
if (rc) {
|
2017-06-03 07:38:06 +00:00
|
|
|
bio->bi_status = rc;
|
2016-01-06 20:03:41 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2015-05-16 16:28:53 +00:00
|
|
|
if (do_acct)
|
2020-05-27 05:24:10 +00:00
|
|
|
bio_end_io_acct(bio, start);
|
2015-06-25 07:08:39 +00:00
|
|
|
|
2016-08-05 21:35:16 +00:00
|
|
|
if (bio->bi_opf & REQ_FUA)
|
2019-07-05 14:03:22 +00:00
|
|
|
ret = nvdimm_flush(nd_region, bio);
|
|
|
|
|
|
|
|
if (ret)
|
|
|
|
bio->bi_status = errno_to_blk_status(ret);
|
2015-06-25 07:08:39 +00:00
|
|
|
|
2015-07-20 13:29:37 +00:00
|
|
|
bio_endio(bio);
|
2015-11-05 17:41:16 +00:00
|
|
|
return BLK_QC_T_NONE;
|
2015-04-01 07:12:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int pmem_rw_page(struct block_device *bdev, sector_t sector,
|
2018-07-18 11:47:36 +00:00
|
|
|
struct page *page, unsigned int op)
|
2015-04-01 07:12:19 +00:00
|
|
|
{
|
2020-05-08 16:15:17 +00:00
|
|
|
struct pmem_device *pmem = bdev->bd_disk->private_data;
|
2017-06-03 07:38:06 +00:00
|
|
|
blk_status_t rc;
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2020-02-28 16:34:51 +00:00
|
|
|
if (op_is_write(op))
|
2020-08-15 00:30:33 +00:00
|
|
|
rc = pmem_do_write(pmem, page, 0, sector, thp_size(page));
|
2020-02-28 16:34:51 +00:00
|
|
|
else
|
2020-08-15 00:30:33 +00:00
|
|
|
rc = pmem_do_read(pmem, page, 0, sector, thp_size(page));
|
2016-01-06 20:03:41 +00:00
|
|
|
/*
|
|
|
|
* The ->rw_page interface is subtle and tricky. The core
|
|
|
|
* retries on any error, so we can only invoke page_endio() in
|
|
|
|
* the successful completion case. Otherwise, we'll see crashes
|
|
|
|
* caused by double completion.
|
|
|
|
*/
|
|
|
|
if (rc == 0)
|
2018-07-18 11:47:36 +00:00
|
|
|
page_endio(page, op_is_write(op), 0);
|
2016-01-06 20:03:41 +00:00
|
|
|
|
2017-06-03 07:38:06 +00:00
|
|
|
return blk_status_to_errno(rc);
|
2015-04-01 07:12:19 +00:00
|
|
|
}
|
|
|
|
|
2016-06-17 18:08:06 +00:00
|
|
|
/* see "strong" declaration in tools/testing/nvdimm/pmem-dax.c */
|
2017-01-25 07:02:09 +00:00
|
|
|
__weak long __pmem_direct_access(struct pmem_device *pmem, pgoff_t pgoff,
|
|
|
|
long nr_pages, void **kaddr, pfn_t *pfn)
|
2015-04-01 07:12:19 +00:00
|
|
|
{
|
2017-01-25 07:02:09 +00:00
|
|
|
resource_size_t offset = PFN_PHYS(pgoff) + pmem->data_offset;
|
2015-10-25 02:55:58 +00:00
|
|
|
|
2017-01-25 07:02:09 +00:00
|
|
|
if (unlikely(is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512,
|
|
|
|
PFN_PHYS(nr_pages))))
|
2016-02-24 22:02:11 +00:00
|
|
|
return -EIO;
|
2018-07-30 07:15:43 +00:00
|
|
|
|
|
|
|
if (kaddr)
|
|
|
|
*kaddr = pmem->virt_addr + offset;
|
|
|
|
if (pfn)
|
|
|
|
*pfn = phys_to_pfn_t(pmem->phys_addr + offset, pmem->pfn_flags);
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2016-02-24 22:02:11 +00:00
|
|
|
/*
|
|
|
|
* If badblocks are present, limit known good range to the
|
|
|
|
* requested range.
|
|
|
|
*/
|
|
|
|
if (unlikely(pmem->bb.count))
|
2017-01-25 07:02:09 +00:00
|
|
|
return nr_pages;
|
|
|
|
return PHYS_PFN(pmem->size - pmem->pfn_pad - offset);
|
2015-04-01 07:12:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static const struct block_device_operations pmem_fops = {
|
|
|
|
.owner = THIS_MODULE,
|
2020-07-01 08:59:43 +00:00
|
|
|
.submit_bio = pmem_submit_bio,
|
2015-04-01 07:12:19 +00:00
|
|
|
.rw_page = pmem_rw_page,
|
|
|
|
};
|
|
|
|
|
2020-02-28 16:34:52 +00:00
|
|
|
static int pmem_dax_zero_page_range(struct dax_device *dax_dev, pgoff_t pgoff,
|
|
|
|
size_t nr_pages)
|
|
|
|
{
|
|
|
|
struct pmem_device *pmem = dax_get_private(dax_dev);
|
|
|
|
|
|
|
|
return blk_status_to_errno(pmem_do_write(pmem, ZERO_PAGE(0), 0,
|
|
|
|
PFN_PHYS(pgoff) >> SECTOR_SHIFT,
|
|
|
|
PAGE_SIZE));
|
|
|
|
}
|
|
|
|
|
2017-01-25 07:02:09 +00:00
|
|
|
static long pmem_dax_direct_access(struct dax_device *dax_dev,
|
|
|
|
pgoff_t pgoff, long nr_pages, void **kaddr, pfn_t *pfn)
|
|
|
|
{
|
|
|
|
struct pmem_device *pmem = dax_get_private(dax_dev);
|
|
|
|
|
|
|
|
return __pmem_direct_access(pmem, pgoff, nr_pages, kaddr, pfn);
|
|
|
|
}
|
|
|
|
|
2019-05-17 00:05:21 +00:00
|
|
|
/*
|
|
|
|
* Use the 'no check' versions of copy_from_iter_flushcache() and
|
x86, powerpc: Rename memcpy_mcsafe() to copy_mc_to_{user, kernel}()
In reaction to a proposal to introduce a memcpy_mcsafe_fast()
implementation Linus points out that memcpy_mcsafe() is poorly named
relative to communicating the scope of the interface. Specifically what
addresses are valid to pass as source, destination, and what faults /
exceptions are handled.
Of particular concern is that even though x86 might be able to handle
the semantics of copy_mc_to_user() with its common copy_user_generic()
implementation other archs likely need / want an explicit path for this
case:
On Fri, May 1, 2020 at 11:28 AM Linus Torvalds <torvalds@linux-foundation.org> wrote:
>
> On Thu, Apr 30, 2020 at 6:21 PM Dan Williams <dan.j.williams@intel.com> wrote:
> >
> > However now I see that copy_user_generic() works for the wrong reason.
> > It works because the exception on the source address due to poison
> > looks no different than a write fault on the user address to the
> > caller, it's still just a short copy. So it makes copy_to_user() work
> > for the wrong reason relative to the name.
>
> Right.
>
> And it won't work that way on other architectures. On x86, we have a
> generic function that can take faults on either side, and we use it
> for both cases (and for the "in_user" case too), but that's an
> artifact of the architecture oddity.
>
> In fact, it's probably wrong even on x86 - because it can hide bugs -
> but writing those things is painful enough that everybody prefers
> having just one function.
Replace a single top-level memcpy_mcsafe() with either
copy_mc_to_user(), or copy_mc_to_kernel().
Introduce an x86 copy_mc_fragile() name as the rename for the
low-level x86 implementation formerly named memcpy_mcsafe(). It is used
as the slow / careful backend that is supplanted by a fast
copy_mc_generic() in a follow-on patch.
One side-effect of this reorganization is that separating copy_mc_64.S
to its own file means that perf no longer needs to track dependencies
for its memcpy_64.S benchmarks.
[ bp: Massage a bit. ]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Cc: <stable@vger.kernel.org>
Link: http://lore.kernel.org/r/CAHk-=wjSqtXAqfUJxFtWNwmguFASTgB0dz1dT3V-78Quiezqbg@mail.gmail.com
Link: https://lkml.kernel.org/r/160195561680.2163339.11574962055305783722.stgit@dwillia2-desk3.amr.corp.intel.com
2020-10-06 03:40:16 +00:00
|
|
|
* copy_mc_to_iter() to bypass HARDENED_USERCOPY overhead. Bounds
|
2019-05-17 00:05:21 +00:00
|
|
|
* checking, both file offset and device offset, is handled by
|
|
|
|
* dax_iomap_actor()
|
|
|
|
*/
|
2017-05-29 19:22:50 +00:00
|
|
|
static size_t pmem_copy_from_iter(struct dax_device *dax_dev, pgoff_t pgoff,
|
|
|
|
void *addr, size_t bytes, struct iov_iter *i)
|
|
|
|
{
|
2019-05-17 00:05:21 +00:00
|
|
|
return _copy_from_iter_flushcache(addr, bytes, i);
|
2017-05-29 19:22:50 +00:00
|
|
|
}
|
|
|
|
|
2018-05-02 13:46:33 +00:00
|
|
|
static size_t pmem_copy_to_iter(struct dax_device *dax_dev, pgoff_t pgoff,
|
|
|
|
void *addr, size_t bytes, struct iov_iter *i)
|
|
|
|
{
|
x86, powerpc: Rename memcpy_mcsafe() to copy_mc_to_{user, kernel}()
In reaction to a proposal to introduce a memcpy_mcsafe_fast()
implementation Linus points out that memcpy_mcsafe() is poorly named
relative to communicating the scope of the interface. Specifically what
addresses are valid to pass as source, destination, and what faults /
exceptions are handled.
Of particular concern is that even though x86 might be able to handle
the semantics of copy_mc_to_user() with its common copy_user_generic()
implementation other archs likely need / want an explicit path for this
case:
On Fri, May 1, 2020 at 11:28 AM Linus Torvalds <torvalds@linux-foundation.org> wrote:
>
> On Thu, Apr 30, 2020 at 6:21 PM Dan Williams <dan.j.williams@intel.com> wrote:
> >
> > However now I see that copy_user_generic() works for the wrong reason.
> > It works because the exception on the source address due to poison
> > looks no different than a write fault on the user address to the
> > caller, it's still just a short copy. So it makes copy_to_user() work
> > for the wrong reason relative to the name.
>
> Right.
>
> And it won't work that way on other architectures. On x86, we have a
> generic function that can take faults on either side, and we use it
> for both cases (and for the "in_user" case too), but that's an
> artifact of the architecture oddity.
>
> In fact, it's probably wrong even on x86 - because it can hide bugs -
> but writing those things is painful enough that everybody prefers
> having just one function.
Replace a single top-level memcpy_mcsafe() with either
copy_mc_to_user(), or copy_mc_to_kernel().
Introduce an x86 copy_mc_fragile() name as the rename for the
low-level x86 implementation formerly named memcpy_mcsafe(). It is used
as the slow / careful backend that is supplanted by a fast
copy_mc_generic() in a follow-on patch.
One side-effect of this reorganization is that separating copy_mc_64.S
to its own file means that perf no longer needs to track dependencies
for its memcpy_64.S benchmarks.
[ bp: Massage a bit. ]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Cc: <stable@vger.kernel.org>
Link: http://lore.kernel.org/r/CAHk-=wjSqtXAqfUJxFtWNwmguFASTgB0dz1dT3V-78Quiezqbg@mail.gmail.com
Link: https://lkml.kernel.org/r/160195561680.2163339.11574962055305783722.stgit@dwillia2-desk3.amr.corp.intel.com
2020-10-06 03:40:16 +00:00
|
|
|
return _copy_mc_to_iter(addr, bytes, i);
|
2018-05-02 13:46:33 +00:00
|
|
|
}
|
|
|
|
|
2017-01-25 07:02:09 +00:00
|
|
|
static const struct dax_operations pmem_dax_ops = {
|
|
|
|
.direct_access = pmem_dax_direct_access,
|
2019-05-16 20:26:29 +00:00
|
|
|
.dax_supported = generic_fsdax_supported,
|
2017-05-29 19:22:50 +00:00
|
|
|
.copy_from_iter = pmem_copy_from_iter,
|
2018-05-02 13:46:33 +00:00
|
|
|
.copy_to_iter = pmem_copy_to_iter,
|
2020-02-28 16:34:52 +00:00
|
|
|
.zero_page_range = pmem_dax_zero_page_range,
|
2017-01-25 07:02:09 +00:00
|
|
|
};
|
|
|
|
|
2017-06-27 04:28:41 +00:00
|
|
|
static const struct attribute_group *pmem_attribute_groups[] = {
|
|
|
|
&dax_attribute_group,
|
|
|
|
NULL,
|
2017-01-25 07:02:09 +00:00
|
|
|
};
|
|
|
|
|
2019-06-26 12:27:09 +00:00
|
|
|
static void pmem_pagemap_cleanup(struct dev_pagemap *pgmap)
|
2016-03-18 03:24:31 +00:00
|
|
|
{
|
2019-06-26 12:27:09 +00:00
|
|
|
struct request_queue *q =
|
|
|
|
container_of(pgmap->ref, struct request_queue, q_usage_counter);
|
2019-06-13 22:56:33 +00:00
|
|
|
|
2016-03-18 03:24:31 +00:00
|
|
|
blk_cleanup_queue(q);
|
|
|
|
}
|
|
|
|
|
2019-06-26 12:27:09 +00:00
|
|
|
static void pmem_release_queue(void *pgmap)
|
2019-06-13 22:56:33 +00:00
|
|
|
{
|
2019-06-26 12:27:09 +00:00
|
|
|
pmem_pagemap_cleanup(pgmap);
|
2019-06-13 22:56:33 +00:00
|
|
|
}
|
|
|
|
|
2019-06-26 12:27:09 +00:00
|
|
|
static void pmem_pagemap_kill(struct dev_pagemap *pgmap)
|
2017-04-28 17:23:37 +00:00
|
|
|
{
|
2019-06-26 12:27:09 +00:00
|
|
|
struct request_queue *q =
|
|
|
|
container_of(pgmap->ref, struct request_queue, q_usage_counter);
|
mm, devm_memremap_pages: fix shutdown handling
The last step before devm_memremap_pages() returns success is to allocate
a release action, devm_memremap_pages_release(), to tear the entire setup
down. However, the result from devm_add_action() is not checked.
Checking the error from devm_add_action() is not enough. The api
currently relies on the fact that the percpu_ref it is using is killed by
the time the devm_memremap_pages_release() is run. Rather than continue
this awkward situation, offload the responsibility of killing the
percpu_ref to devm_memremap_pages_release() directly. This allows
devm_memremap_pages() to do the right thing relative to init failures and
shutdown.
Without this change we could fail to register the teardown of
devm_memremap_pages(). The likelihood of hitting this failure is tiny as
small memory allocations almost always succeed. However, the impact of
the failure is large given any future reconfiguration, or disable/enable,
of an nvdimm namespace will fail forever as subsequent calls to
devm_memremap_pages() will fail to setup the pgmap_radix since there will
be stale entries for the physical address range.
An argument could be made to require that the ->kill() operation be set in
the @pgmap arg rather than passed in separately. However, it helps code
readability, tracking the lifetime of a given instance, to be able to grep
the kill routine directly at the devm_memremap_pages() call site.
Link: http://lkml.kernel.org/r/154275558526.76910.7535251937849268605.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Fixes: e8d513483300 ("memremap: change devm_memremap_pages interface...")
Reviewed-by: "Jérôme Glisse" <jglisse@redhat.com>
Reported-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 08:34:57 +00:00
|
|
|
|
2017-05-02 06:54:56 +00:00
|
|
|
blk_freeze_queue_start(q);
|
2017-04-28 17:23:37 +00:00
|
|
|
}
|
|
|
|
|
2017-01-25 07:02:09 +00:00
|
|
|
static void pmem_release_disk(void *__pmem)
|
2016-03-18 03:24:31 +00:00
|
|
|
{
|
2017-01-25 07:02:09 +00:00
|
|
|
struct pmem_device *pmem = __pmem;
|
|
|
|
|
|
|
|
kill_dax(pmem->dax_dev);
|
|
|
|
put_dax(pmem->dax_dev);
|
|
|
|
del_gendisk(pmem->disk);
|
|
|
|
put_disk(pmem->disk);
|
2016-03-18 03:24:31 +00:00
|
|
|
}
|
|
|
|
|
2019-06-26 12:27:08 +00:00
|
|
|
static const struct dev_pagemap_ops fsdax_pagemap_ops = {
|
|
|
|
.kill = pmem_pagemap_kill,
|
|
|
|
.cleanup = pmem_pagemap_cleanup,
|
|
|
|
};
|
|
|
|
|
2016-03-22 07:22:16 +00:00
|
|
|
static int pmem_attach_disk(struct device *dev,
|
|
|
|
struct nd_namespace_common *ndns)
|
2015-04-01 07:12:19 +00:00
|
|
|
{
|
2016-03-22 07:22:16 +00:00
|
|
|
struct nd_namespace_io *nsio = to_nd_namespace_io(&ndns->dev);
|
2016-07-08 02:44:50 +00:00
|
|
|
struct nd_region *nd_region = to_nd_region(dev->parent);
|
2018-06-06 16:45:13 +00:00
|
|
|
int nid = dev_to_node(dev), fua;
|
2016-03-22 07:22:16 +00:00
|
|
|
struct resource *res = &nsio->res;
|
2017-12-29 07:54:05 +00:00
|
|
|
struct resource bb_res;
|
2016-03-22 07:22:16 +00:00
|
|
|
struct nd_pfn *nd_pfn = NULL;
|
2017-01-25 07:02:09 +00:00
|
|
|
struct dax_device *dax_dev;
|
2016-03-22 07:22:16 +00:00
|
|
|
struct nd_pfn_sb *pfn_sb;
|
2015-04-01 07:12:19 +00:00
|
|
|
struct pmem_device *pmem;
|
2016-01-16 00:56:46 +00:00
|
|
|
struct request_queue *q;
|
2017-06-27 04:28:41 +00:00
|
|
|
struct device *gendev;
|
2016-03-22 07:22:16 +00:00
|
|
|
struct gendisk *disk;
|
|
|
|
void *addr;
|
2017-12-29 07:54:05 +00:00
|
|
|
int rc;
|
2019-07-05 14:03:24 +00:00
|
|
|
unsigned long flags = 0UL;
|
2017-12-29 07:54:05 +00:00
|
|
|
|
|
|
|
pmem = devm_kzalloc(dev, sizeof(*pmem), GFP_KERNEL);
|
|
|
|
if (!pmem)
|
|
|
|
return -ENOMEM;
|
2016-03-22 07:22:16 +00:00
|
|
|
|
2019-10-31 10:57:41 +00:00
|
|
|
rc = devm_namespace_enable(dev, ndns, nd_info_block_reserve());
|
|
|
|
if (rc)
|
|
|
|
return rc;
|
|
|
|
|
2016-03-22 07:22:16 +00:00
|
|
|
/* while nsio_rw_bytes is active, parse a pfn info block if present */
|
|
|
|
if (is_nd_pfn(dev)) {
|
|
|
|
nd_pfn = to_nd_pfn(dev);
|
2017-12-29 07:54:05 +00:00
|
|
|
rc = nvdimm_setup_pfn(nd_pfn, &pmem->pgmap);
|
|
|
|
if (rc)
|
|
|
|
return rc;
|
2016-03-22 07:22:16 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* we're attaching a block device, disable raw namespace access */
|
2019-10-31 10:57:41 +00:00
|
|
|
devm_namespace_disable(dev, ndns);
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2016-03-22 07:22:16 +00:00
|
|
|
dev_set_drvdata(dev, pmem);
|
2015-04-01 07:12:19 +00:00
|
|
|
pmem->phys_addr = res->start;
|
|
|
|
pmem->size = resource_size(res);
|
2017-06-09 16:46:50 +00:00
|
|
|
fua = nvdimm_has_flush(nd_region);
|
|
|
|
if (!IS_ENABLED(CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE) || fua < 0) {
|
2015-06-25 07:08:39 +00:00
|
|
|
dev_warn(dev, "unable to guarantee persistence of writes\n");
|
2017-06-09 16:46:50 +00:00
|
|
|
fua = 0;
|
|
|
|
}
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2016-03-22 05:28:40 +00:00
|
|
|
if (!devm_request_mem_region(dev, res->start, resource_size(res),
|
libnvdimm: use consistent naming for request_mem_region()
Here is an example /proc/iomem listing for a system with 2 namespaces,
one in "sector" mode and one in "memory" mode:
1fc000000-2fbffffff : Persistent Memory (legacy)
1fc000000-2fbffffff : namespace1.0
340000000-34fffffff : Persistent Memory
340000000-34fffffff : btt0.1
Here is the corresponding ndctl listing:
# ndctl list
[
{
"dev":"namespace1.0",
"mode":"memory",
"size":4294967296,
"blockdev":"pmem1"
},
{
"dev":"namespace0.0",
"mode":"sector",
"size":267091968,
"uuid":"f7594f86-badb-4592-875f-ded577da2eaf",
"sector_size":4096,
"blockdev":"pmem0s"
}
]
Notice that the ndctl listing is purely in terms of namespace devices,
while the iomem listing leaks the internal "btt0.1" implementation
detail. Given that ndctl requires the namespace device name to change
the mode, for example:
# ndctl create-namespace --reconfig=namespace0.0 --mode=raw --force
...use the namespace name in the iomem listing to keep the claiming
device name consistent across different mode settings.
Cc: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2016-11-28 19:15:18 +00:00
|
|
|
dev_name(&ndns->dev))) {
|
2016-03-22 05:28:40 +00:00
|
|
|
dev_warn(dev, "could not reserve region %pR\n", res);
|
2016-03-22 07:22:16 +00:00
|
|
|
return -EBUSY;
|
2015-04-01 07:12:19 +00:00
|
|
|
}
|
|
|
|
|
2020-07-01 08:59:43 +00:00
|
|
|
q = blk_alloc_queue(dev_to_node(dev));
|
2016-01-16 00:56:46 +00:00
|
|
|
if (!q)
|
2016-03-22 07:22:16 +00:00
|
|
|
return -ENOMEM;
|
2016-01-16 00:56:46 +00:00
|
|
|
|
2016-01-16 00:56:14 +00:00
|
|
|
pmem->pfn_flags = PFN_DEV;
|
2017-12-29 07:54:05 +00:00
|
|
|
pmem->pgmap.ref = &q->q_usage_counter;
|
2016-03-22 07:22:16 +00:00
|
|
|
if (is_nd_pfn(dev)) {
|
2019-06-26 12:27:10 +00:00
|
|
|
pmem->pgmap.type = MEMORY_DEVICE_FS_DAX;
|
|
|
|
pmem->pgmap.ops = &fsdax_pagemap_ops;
|
2017-12-29 07:54:05 +00:00
|
|
|
addr = devm_memremap_pages(dev, &pmem->pgmap);
|
2016-03-22 07:22:16 +00:00
|
|
|
pfn_sb = nd_pfn->pfn_sb;
|
|
|
|
pmem->data_offset = le64_to_cpu(pfn_sb->dataoff);
|
2017-12-29 07:54:05 +00:00
|
|
|
pmem->pfn_pad = resource_size(res) -
|
|
|
|
resource_size(&pmem->pgmap.res);
|
2016-03-22 07:22:16 +00:00
|
|
|
pmem->pfn_flags |= PFN_MAP;
|
2017-12-29 07:54:05 +00:00
|
|
|
memcpy(&bb_res, &pmem->pgmap.res, sizeof(bb_res));
|
|
|
|
bb_res.start += pmem->data_offset;
|
2016-03-22 07:22:16 +00:00
|
|
|
} else if (pmem_should_map_pages(dev)) {
|
2017-12-29 07:54:05 +00:00
|
|
|
memcpy(&pmem->pgmap.res, &nsio->res, sizeof(pmem->pgmap.res));
|
2019-06-26 12:27:10 +00:00
|
|
|
pmem->pgmap.type = MEMORY_DEVICE_FS_DAX;
|
|
|
|
pmem->pgmap.ops = &fsdax_pagemap_ops;
|
2017-12-29 07:54:05 +00:00
|
|
|
addr = devm_memremap_pages(dev, &pmem->pgmap);
|
2016-01-16 00:56:14 +00:00
|
|
|
pmem->pfn_flags |= PFN_MAP;
|
2017-12-29 07:54:05 +00:00
|
|
|
memcpy(&bb_res, &pmem->pgmap.res, sizeof(bb_res));
|
2018-10-04 23:32:08 +00:00
|
|
|
} else {
|
2019-06-13 22:56:33 +00:00
|
|
|
if (devm_add_action_or_reset(dev, pmem_release_queue,
|
2019-06-26 12:27:09 +00:00
|
|
|
&pmem->pgmap))
|
2019-06-13 22:56:33 +00:00
|
|
|
return -ENOMEM;
|
2016-03-22 07:22:16 +00:00
|
|
|
addr = devm_memremap(dev, pmem->phys_addr,
|
|
|
|
pmem->size, ARCH_MEMREMAP_PMEM);
|
2018-10-04 23:32:08 +00:00
|
|
|
memcpy(&bb_res, &nsio->res, sizeof(bb_res));
|
|
|
|
}
|
2015-09-15 06:42:20 +00:00
|
|
|
|
2016-03-22 07:22:16 +00:00
|
|
|
if (IS_ERR(addr))
|
|
|
|
return PTR_ERR(addr);
|
2016-06-04 01:06:47 +00:00
|
|
|
pmem->virt_addr = addr;
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2018-06-06 16:45:13 +00:00
|
|
|
blk_queue_write_cache(q, true, fua);
|
2016-03-21 22:43:53 +00:00
|
|
|
blk_queue_physical_block_size(q, PAGE_SIZE);
|
2017-06-04 03:12:07 +00:00
|
|
|
blk_queue_logical_block_size(q, pmem_sector_size(ndns));
|
2016-03-21 22:43:53 +00:00
|
|
|
blk_queue_max_hw_sectors(q, UINT_MAX);
|
2018-03-08 01:10:10 +00:00
|
|
|
blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
|
2018-06-26 22:30:39 +00:00
|
|
|
if (pmem->pfn_flags & PFN_MAP)
|
|
|
|
blk_queue_flag_set(QUEUE_FLAG_DAX, q);
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2015-10-06 00:35:56 +00:00
|
|
|
disk = alloc_disk_node(0, nid);
|
2016-03-18 03:24:31 +00:00
|
|
|
if (!disk)
|
|
|
|
return -ENOMEM;
|
2017-01-25 07:02:09 +00:00
|
|
|
pmem->disk = disk;
|
2015-04-01 07:12:19 +00:00
|
|
|
|
|
|
|
disk->fops = &pmem_fops;
|
2016-03-21 22:43:53 +00:00
|
|
|
disk->queue = q;
|
2015-04-01 07:12:19 +00:00
|
|
|
disk->flags = GENHD_FL_EXT_DEVT;
|
2020-05-08 16:15:17 +00:00
|
|
|
disk->private_data = pmem;
|
nd_btt: atomic sector updates
BTT stands for Block Translation Table, and is a way to provide power
fail sector atomicity semantics for block devices that have the ability
to perform byte granularity IO. It relies on the capability of libnvdimm
namespace devices to do byte aligned IO.
The BTT works as a stacked blocked device, and reserves a chunk of space
from the backing device for its accounting metadata. It is a bio-based
driver because all IO is done synchronously, and there is no queuing or
asynchronous completions at either the device or the driver level.
The BTT uses 'lanes' to index into various 'on-disk' data structures,
and lanes also act as a synchronization mechanism in case there are more
CPUs than available lanes. We did a comparison between two lane lock
strategies - first where we kept an atomic counter around that tracked
which was the last lane that was used, and 'our' lane was determined by
atomically incrementing that. That way, for the nr_cpus > nr_lanes case,
theoretically, no CPU would be blocked waiting for a lane. The other
strategy was to use the cpu number we're scheduled on to and hash it to
a lane number. Theoretically, this could block an IO that could've
otherwise run using a different, free lane. But some fio workloads
showed that the direct cpu -> lane hash performed faster than tracking
'last lane' - my reasoning is the cache thrash caused by moving the
atomic variable made that approach slower than simply waiting out the
in-progress IO. This supports the conclusion that the driver can be a
very simple bio-based one that does synchronous IOs instead of queuing.
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Boaz Harrosh <boaz@plexistor.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jens Axboe <axboe@fb.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Neil Brown <neilb@suse.de>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
[jmoyer: fix nmi watchdog timeout in btt_map_init]
[jmoyer: move btt initialization to module load path]
[jmoyer: fix memory leak in the btt initialization path]
[jmoyer: Don't overwrite corrupted arenas]
Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2015-06-25 08:20:32 +00:00
|
|
|
nvdimm_namespace_disk_name(ndns, disk->disk_name);
|
2016-03-03 17:38:00 +00:00
|
|
|
set_capacity(disk, (pmem->size - pmem->pfn_pad - pmem->data_offset)
|
|
|
|
/ 512);
|
2016-01-05 07:50:23 +00:00
|
|
|
if (devm_init_badblocks(dev, &pmem->bb))
|
|
|
|
return -ENOMEM;
|
2017-12-29 07:54:05 +00:00
|
|
|
nvdimm_badblocks_populate(nd_region, &pmem->bb, &bb_res);
|
2016-01-06 20:03:42 +00:00
|
|
|
disk->bb = &pmem->bb;
|
2016-06-15 21:59:17 +00:00
|
|
|
|
2019-07-05 14:03:24 +00:00
|
|
|
if (is_nvdimm_sync(nd_region))
|
|
|
|
flags = DAXDEV_F_SYNC;
|
|
|
|
dax_dev = alloc_dax(pmem, disk->disk_name, &pmem_dax_ops, flags);
|
2020-04-01 16:11:25 +00:00
|
|
|
if (IS_ERR(dax_dev)) {
|
2017-01-25 07:02:09 +00:00
|
|
|
put_disk(disk);
|
2020-04-01 16:11:25 +00:00
|
|
|
return PTR_ERR(dax_dev);
|
2017-01-25 07:02:09 +00:00
|
|
|
}
|
2018-06-06 16:45:13 +00:00
|
|
|
dax_write_cache(dax_dev, nvdimm_has_cache(nd_region));
|
2017-01-25 07:02:09 +00:00
|
|
|
pmem->dax_dev = dax_dev;
|
2017-06-27 04:28:41 +00:00
|
|
|
gendev = disk_to_dev(disk);
|
|
|
|
gendev->groups = pmem_attribute_groups;
|
|
|
|
|
2018-09-28 06:17:19 +00:00
|
|
|
device_add_disk(dev, disk, NULL);
|
2017-01-25 07:02:09 +00:00
|
|
|
if (devm_add_action_or_reset(dev, pmem_release_disk, pmem))
|
2016-06-15 21:59:17 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
|
2020-09-01 15:57:47 +00:00
|
|
|
nvdimm_check_and_set_ro(disk);
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2017-06-12 22:25:11 +00:00
|
|
|
pmem->bb_state = sysfs_get_dirent(disk_to_dev(disk)->kobj.sd,
|
|
|
|
"badblocks");
|
2017-07-01 01:56:03 +00:00
|
|
|
if (!pmem->bb_state)
|
|
|
|
dev_warn(dev, "'badblocks' notification disabled\n");
|
2017-06-12 22:25:11 +00:00
|
|
|
|
2015-06-25 08:20:04 +00:00
|
|
|
return 0;
|
|
|
|
}
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2015-06-09 19:33:45 +00:00
|
|
|
static int nd_pmem_probe(struct device *dev)
|
2015-04-01 07:12:19 +00:00
|
|
|
{
|
2019-09-05 15:45:58 +00:00
|
|
|
int ret;
|
2015-06-25 08:20:04 +00:00
|
|
|
struct nd_namespace_common *ndns;
|
2015-04-01 07:12:19 +00:00
|
|
|
|
2015-06-25 08:20:04 +00:00
|
|
|
ndns = nvdimm_namespace_common_probe(dev);
|
|
|
|
if (IS_ERR(ndns))
|
|
|
|
return PTR_ERR(ndns);
|
2015-06-17 21:14:46 +00:00
|
|
|
|
2016-03-22 07:22:16 +00:00
|
|
|
if (is_nd_btt(dev))
|
2015-08-11 03:07:08 +00:00
|
|
|
return nvdimm_namespace_attach_btt(ndns);
|
|
|
|
|
2015-08-01 06:16:37 +00:00
|
|
|
if (is_nd_pfn(dev))
|
2016-03-22 07:22:16 +00:00
|
|
|
return pmem_attach_disk(dev, ndns);
|
2015-08-01 06:16:37 +00:00
|
|
|
|
2019-10-31 10:57:41 +00:00
|
|
|
ret = devm_namespace_enable(dev, ndns, nd_info_block_reserve());
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
2019-09-05 15:45:58 +00:00
|
|
|
ret = nd_btt_probe(dev, ndns);
|
|
|
|
if (ret == 0)
|
2015-08-01 06:16:37 +00:00
|
|
|
return -ENXIO;
|
|
|
|
|
2019-09-05 15:45:58 +00:00
|
|
|
/*
|
|
|
|
* We have two failure conditions here, there is no
|
|
|
|
* info reserver block or we found a valid info reserve block
|
|
|
|
* but failed to initialize the pfn superblock.
|
|
|
|
*
|
|
|
|
* For the first case consider namespace as a raw pmem namespace
|
|
|
|
* and attach a disk.
|
|
|
|
*
|
|
|
|
* For the latter, consider this a success and advance the namespace
|
|
|
|
* seed.
|
|
|
|
*/
|
|
|
|
ret = nd_pfn_probe(dev, ndns);
|
|
|
|
if (ret == 0)
|
|
|
|
return -ENXIO;
|
|
|
|
else if (ret == -EOPNOTSUPP)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = nd_dax_probe(dev, ndns);
|
|
|
|
if (ret == 0)
|
|
|
|
return -ENXIO;
|
|
|
|
else if (ret == -EOPNOTSUPP)
|
|
|
|
return ret;
|
2019-10-31 10:57:41 +00:00
|
|
|
|
|
|
|
/* probe complete, attach handles namespace enabling */
|
|
|
|
devm_namespace_disable(dev, ndns);
|
|
|
|
|
2016-03-22 07:22:16 +00:00
|
|
|
return pmem_attach_disk(dev, ndns);
|
2015-04-01 07:12:19 +00:00
|
|
|
}
|
|
|
|
|
2015-06-09 19:33:45 +00:00
|
|
|
static int nd_pmem_remove(struct device *dev)
|
2015-04-01 07:12:19 +00:00
|
|
|
{
|
2017-07-01 01:56:03 +00:00
|
|
|
struct pmem_device *pmem = dev_get_drvdata(dev);
|
|
|
|
|
2015-06-25 08:20:04 +00:00
|
|
|
if (is_nd_btt(dev))
|
2016-03-15 23:41:04 +00:00
|
|
|
nvdimm_namespace_detach_btt(to_nd_btt(dev));
|
2017-07-01 01:56:03 +00:00
|
|
|
else {
|
|
|
|
/*
|
2019-07-18 01:08:26 +00:00
|
|
|
* Note, this assumes nd_device_lock() context to not
|
|
|
|
* race nd_pmem_notify()
|
2017-07-01 01:56:03 +00:00
|
|
|
*/
|
|
|
|
sysfs_put(pmem->bb_state);
|
|
|
|
pmem->bb_state = NULL;
|
|
|
|
}
|
2019-07-05 14:03:22 +00:00
|
|
|
nvdimm_flush(to_nd_region(dev->parent), NULL);
|
2016-07-09 07:12:52 +00:00
|
|
|
|
2015-04-01 07:12:19 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2016-07-09 07:12:52 +00:00
|
|
|
static void nd_pmem_shutdown(struct device *dev)
|
|
|
|
{
|
2019-07-05 14:03:22 +00:00
|
|
|
nvdimm_flush(to_nd_region(dev->parent), NULL);
|
2016-07-09 07:12:52 +00:00
|
|
|
}
|
|
|
|
|
2016-02-18 18:29:49 +00:00
|
|
|
static void nd_pmem_notify(struct device *dev, enum nvdimm_event event)
|
|
|
|
{
|
2017-04-25 23:04:13 +00:00
|
|
|
struct nd_region *nd_region;
|
2016-03-15 23:41:04 +00:00
|
|
|
resource_size_t offset = 0, end_trunc = 0;
|
|
|
|
struct nd_namespace_common *ndns;
|
|
|
|
struct nd_namespace_io *nsio;
|
|
|
|
struct resource res;
|
2017-04-25 23:04:13 +00:00
|
|
|
struct badblocks *bb;
|
2017-06-12 22:25:11 +00:00
|
|
|
struct kernfs_node *bb_state;
|
2016-02-18 18:29:49 +00:00
|
|
|
|
|
|
|
if (event != NVDIMM_REVALIDATE_POISON)
|
|
|
|
return;
|
|
|
|
|
2016-03-15 23:41:04 +00:00
|
|
|
if (is_nd_btt(dev)) {
|
|
|
|
struct nd_btt *nd_btt = to_nd_btt(dev);
|
|
|
|
|
|
|
|
ndns = nd_btt->ndns;
|
2017-04-25 23:04:13 +00:00
|
|
|
nd_region = to_nd_region(ndns->dev.parent);
|
|
|
|
nsio = to_nd_namespace_io(&ndns->dev);
|
|
|
|
bb = &nsio->bb;
|
2017-06-12 22:25:11 +00:00
|
|
|
bb_state = NULL;
|
2017-04-25 23:04:13 +00:00
|
|
|
} else {
|
|
|
|
struct pmem_device *pmem = dev_get_drvdata(dev);
|
2016-04-08 03:02:06 +00:00
|
|
|
|
2017-04-25 23:04:13 +00:00
|
|
|
nd_region = to_region(pmem);
|
|
|
|
bb = &pmem->bb;
|
2017-06-12 22:25:11 +00:00
|
|
|
bb_state = pmem->bb_state;
|
2017-04-25 23:04:13 +00:00
|
|
|
|
|
|
|
if (is_nd_pfn(dev)) {
|
|
|
|
struct nd_pfn *nd_pfn = to_nd_pfn(dev);
|
|
|
|
struct nd_pfn_sb *pfn_sb = nd_pfn->pfn_sb;
|
|
|
|
|
|
|
|
ndns = nd_pfn->ndns;
|
|
|
|
offset = pmem->data_offset +
|
|
|
|
__le32_to_cpu(pfn_sb->start_pad);
|
|
|
|
end_trunc = __le32_to_cpu(pfn_sb->end_trunc);
|
|
|
|
} else {
|
|
|
|
ndns = to_ndns(dev);
|
|
|
|
}
|
|
|
|
|
|
|
|
nsio = to_nd_namespace_io(&ndns->dev);
|
|
|
|
}
|
2016-04-08 03:02:06 +00:00
|
|
|
|
2016-03-15 23:41:04 +00:00
|
|
|
res.start = nsio->res.start + offset;
|
|
|
|
res.end = nsio->res.end - end_trunc;
|
2017-04-25 23:04:13 +00:00
|
|
|
nvdimm_badblocks_populate(nd_region, bb, &res);
|
2017-06-12 22:25:11 +00:00
|
|
|
if (bb_state)
|
|
|
|
sysfs_notify_dirent(bb_state);
|
2016-02-18 18:29:49 +00:00
|
|
|
}
|
|
|
|
|
2015-06-09 19:33:45 +00:00
|
|
|
MODULE_ALIAS("pmem");
|
|
|
|
MODULE_ALIAS_ND_DEVICE(ND_DEVICE_NAMESPACE_IO);
|
2015-06-17 21:14:46 +00:00
|
|
|
MODULE_ALIAS_ND_DEVICE(ND_DEVICE_NAMESPACE_PMEM);
|
2015-06-09 19:33:45 +00:00
|
|
|
static struct nd_device_driver nd_pmem_driver = {
|
|
|
|
.probe = nd_pmem_probe,
|
|
|
|
.remove = nd_pmem_remove,
|
2016-02-18 18:29:49 +00:00
|
|
|
.notify = nd_pmem_notify,
|
2016-07-09 07:12:52 +00:00
|
|
|
.shutdown = nd_pmem_shutdown,
|
2015-06-09 19:33:45 +00:00
|
|
|
.drv = {
|
|
|
|
.name = "nd_pmem",
|
2015-04-01 07:12:19 +00:00
|
|
|
},
|
2015-06-17 21:14:46 +00:00
|
|
|
.type = ND_DRIVER_NAMESPACE_IO | ND_DRIVER_NAMESPACE_PMEM,
|
2015-04-01 07:12:19 +00:00
|
|
|
};
|
|
|
|
|
2018-03-14 18:25:07 +00:00
|
|
|
module_nd_driver(nd_pmem_driver);
|
2015-04-01 07:12:19 +00:00
|
|
|
|
|
|
|
MODULE_AUTHOR("Ross Zwisler <ross.zwisler@linux.intel.com>");
|
|
|
|
MODULE_LICENSE("GPL v2");
|