>>
>> Most probably,
>>
>> struct range memhp_get_addressable_range(bool need_mapping)
>> {
>> ...
>> }
>
> Something like this...
>
> +struct memhp_range {
> + u64 start;
> + u64 end;
> +};
We do have struct range already in include/linux/range.h
> +
> +#ifndef arch_get_addressable_range
> +static inline struct memhp_range arch_get_mappable_range(bool need_mapping)
> +{
> + struct memhp_range range = {
> + .start = 0UL,
> + .end = (1ull << (MAX_PHYSMEM_BITS + 1)) - 1,
Or just set to -1ULL if it's only used in memhp_get_mappable_range(), to
keep things simple ().
> + };
> + return range;
> +}
> +#endif
> +
> +static inline struct memhp_range memhp_get_mappable_range(bool need_mapping)
due to "need_mapping" the function might better be called
memhp_get_pluggable_range()
or similar
> +{
> + const u64 max_phys = (1ull << (MAX_PHYSMEM_BITS + 1)) - 1;
> + struct memhp_range range = arch_get_mappable_range(need_mapping);
> +
> + if (range.start > max_phys) {
> + range.start = 0;
> + range.end = 0;
> + }
> + range.end = min_t(u64, range.end, max_phys);
> + return range;
> +}
> +
> +static inline bool memhp_range_allowed(u64 start, u64 end, bool need_mapping)
> +{
> + struct memhp_range range = memhp_get_mappable_range(need_mapping);
> +
> + return (start <= end) && (start >= range.start) && (end <= range.end);
Keep in mind that in memory hotplug code, "end" is usually exclusive,
and "end" in "struct range" is inclusive (see range_len(), or how you
calculate max_phys.
So depending on the semantics, you might have to fixup your comparisons.
return start < end && start >= range.start && end <= range.end - 1;
[...]
>> Right now it's like calling a function with wrong arguments - you just
>> don't have a clue what valid arguments are, because non-obvious errors
>> (besides -ENOMEM, which is a temporary error) pop up deep down the call
>> chain.
>>
>> For example, virito-mem would use it to detect during device
>> initialization the usable device range, and warn the user accordingly.
>> It currently manually checks for MAX_PHYSMEM_BITS, but that's just ugly.
>> Failing at random add_memory() calls (permanently!) is not so nice.
>>
>> In case of DIMMs, we could use it to detect if adding parts of a DIMM
>> won't work (and warn the user early). We could try to add as much as
>> possible.
>
> Agreed.
>
> Planning to add memhp_range_allowed() check in add_memory(), __add_memory(),
> add_memory_driver_managed() and memremap_pages(). This check might just get
> called twice depending on the hotplug path. Wondering if this needs to be
> added any where else ?
So
add_memory() needs to
- add sections via arch_add_memory()
- create a mapping via arch_add_memory()->add_pages()
memremap_pages() via arch_add_memory() needs to
- add sections via arch_add_memory()
- create a mapping via arch_add_memory()->add_pages()
memremap_pages() via add_pages() needs to
- add sections
I'll reuse the functions from virtio-mem code once in place (exposing
memhp_get_pluggable_range()).
I do agree that having the callers of arch_add_memory() / add_pages()
validate stuff isn't completely nice. I already raised that I would much
rather want to see !arch wrappers for these arch functions that could
validate stuff. But then we would have to do a bigger cleanup to get
naming right.
1. Rename functions for handling system ram like
s/add_memory/add_sysram/
s/remove_memory/remove_sysram/
...
2. Have a new add_memory() that validates + calls arch_add_memory()
3. s/add_pages/arch_add_pages/
4. Have a new add_pages() that validates + calls arch_add_pages()
...
Long story short, handling it in the 2 (!) callers might be easier for now.
--
Thanks,
David / dhildenb
