git: 3b15beb30b3b - main - Implement malloc_domainset_aligned(9).

[Jessica Clarke]

On 18 Jan 2021, at 21:19, Konstantin Belousov <kostikbel at gmail.com> wrote:
> On Mon, Jan 18, 2021 at 08:45:43PM +0000, Alexander V. Chernikov wrote:
>> 17.01.2021, 17:30, "Konstantin Belousov" <kib at freebsd.org>:
>>> The branch main has been updated by kib:
>>> 
>>> URL: https://cgit.FreeBSD.org/src/commit/?id=3b15beb30b3b4ba17bae3d1d43c8c04ff862bb57
>>> 
>>> commit 3b15beb30b3b4ba17bae3d1d43c8c04ff862bb57
>>> Author: Konstantin Belousov <kib at FreeBSD.org>
>>> AuthorDate: 2021-01-14 03:59:34 +0000
>>> Commit: Konstantin Belousov <kib at FreeBSD.org>
>>> CommitDate: 2021-01-17 17:29:05 +0000
>>> 
>>>     Implement malloc_domainset_aligned(9).
>> Hi Kostik,
>> 
>> This change makes my vm panic in usb code. No dump, as dumpdev not mounted yet.
>> 
>> Note: the below lines have been OCR'ed, so there may be some errors.
>> 
>> Root mount waiting for: CAM usbus0 usbus1
>> panic: malloc_domainset_aligned: result not aligned 0xfffff8000551ca80 size 0x180 align 0x100
> 
> Try this.
> 
> ...
> 	if (size < align)
> -		size = align;
> -	res = malloc_domainset(size, mtp, ds, flags);
> +		asize = align;
> +	else if (!powerof2(size))
> +		asize = roundup2(size, align);
> +	else
> +		asize = size;

Why not just `asize = roundup2(size, align)`? It's only 3 instructions
(4 on Clang due to a missed combine) on amd64 and I'd be very surprised
if the branching version were faster even in the fast path case. Plus
it's much easier to read; currently it looks like there's something
unusual going on but there really isn't.

https://godbolt.org/z/rv6qs7 FWIW

Jess