Re: [PATCH 3/4] rhashtable: use bit_spin_locks to protect hash bucket.

From: NeilBrown
Date: Tue Mar 26 2019 - 21:10:21 EST


On Tue, Mar 26 2019, Herbert Xu wrote:

> On Mon, Mar 25, 2019 at 04:05:39PM +1100, NeilBrown wrote:
>>
>> +/*
>> + * If 'p' is a bucket head and might be locked:
>> + * rht_ptr() returns the address without the lock bit.
>> + * rht_ptr_locked() returns the address WITH the lock bit.
>> + * rht_is_locked() tests if the lock bit is set.
>> + */
>> +static inline struct rhash_head __rcu *rht_ptr(const struct rhash_head *p)
>> +{
>> + return (void *)(((unsigned long)p) & ~2UL);
>> +}
>
> My preference would be to turn rhash_head * into an opaque type
> that can only be dereferenced through rht_ptr.

I can try that.
Presumably the 'next' pointer in 'struct rhash_head' can still be a
normal pointer, but the buckets in the bucket_table would need to
opaque pointers.
Maybe pointers to some new undefined struct, which gets cast and masked
as appropriate?

I'll have a go and see how it looks.

Thanks,
NeilBrown

Attachment: signature.asc
Description: PGP signature