On Wed, 2016-02-17 at 10:29 +0100, Ingo Molnar wrote:
* Toshi Kani <toshi.kani(a)hpe.com> wrote:
> x86 does not define ARCH_HAS_VALID_PHYS_ADDR_RANGE, which
> leads /dev/mem to use the default valid_phys_addr_range()
> and valid_mmap_phys_addr_range() in drivers/char/mem.c.
> The default valid_phys_addr_range() allows any range lower
> than __pa(high_memory), which is the end of system RAM, and
> disallows any range higher than it.
> Persistent memory may be located at lower and/or higher
> address of __pa(high_memory) depending on their memory slots.
> When using crash(8) via /dev/mem for analyzing data in
> persistent memory, it can only access to the one lower than
> Add x86 valid_phys_addr_range() and valid_mmap_phys_addr_range()
> to provide better checking:
> - Physical address range is valid when it is fully backed by
> IORESOURCE_MEM, regardless of __pa(high_memory).
> - Other ranges, including holes, are invalid.
> This also allows crash(8) to access persistent memory ranges
> via /dev/mem (with a minor change to remove high_memory check
> from crash itself).
> Note, /dev/mem makes additional check with devmem_is_allowed()
> for read/write when CONFIG_STRICT_DEVMEM is set, and does always
> for mmap. CONFIG_IO_STRICT_DEVMEM provides further restriction.
So it's hard to judge the quality of these new APIs without
actual usecases. So please Cc: me to whatever work this is used in, and
I'll have a look in that context.
Great! The source code of crash(8) is available in the github below.
crash is a tool to analyze memory data via a crashdump file or /dev/mem. I
am trying to make this tool works on NVDIMM via /dev/mem. (NVDIMM ranges
are not saved to a crashdump file, but are persistent anyway.) We had BTT
metadata corruptions, and this tool can be helpful to verify such data
until we have better tools.
When /dev/mem is specified as the source, read_dev_mem() is set to pc-
readmem as the read function to /dev/mem. read_dev_mem() is at line
of the file blow.
read_dev_mem() has the same high_memory check (i.e. mirroring the /dev/mem
restriction), which should also be removed after this patch is accepted.
Falling back to /dev/kmem does not help since read_kmem() in the
kernel has check to high_memory in the function itself. With this patch,
read_dev_mem() works on NVDIMM ranges.
Let me know if you have any question.