[patch] x86/realmode: Make stack lock work in trampoline_compat()

David Laight David.Laight at ACULAB.COM
Sat Jun 10 12:50:58 PDT 2023


From: Andrew Cooper
> Sent: 09 June 2023 00:58
> 
...
> The important point is the l suffix on btsl, which forces it to be long
> (32bit) irrespective of the default operand size.

Does that matter at all?
The 'bit' opcodes (I'm sure 'bts' is 'bit test and set') take
a bit offset from the base address.
This accesses the same bit regardless of the operand size.

The one real issue is that a byte operand will only lock the one byte.
This might be problematic if non-bit locked accesses are also used.
Although it would need to be rather obscure use.
(This may be one of them...)

The only other problem is that btsl always does a locked 32bit
access. If the base address is misaligned this is a misaligned
locked access - problematic if it crosses a cache line boundary.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)


More information about the linux-riscv mailing list