On 3/29/23 1:42 PM, Jens Axboe wrote:
> On 3/29/23 1:38 PM, Jens Axboe wrote:
>> On 3/29/23 1:30 PM, Linus Torvalds wrote:
>>> On Wed, Mar 29, 2023 at 11:41 AM Jens Axboe <axboe@xxxxxxxxx> wrote:
>>>>
>>>> + struct iovec __ubuf_iovec;
>>>
>>> I think this is the third time I say this: this should be "const struct iovec".
>>
>> Doh sorry, not sure why I keep missing that... But yes, it should, I'll make
>> the edit and actually amend it.
>
> Now I recall why that ended up like that again, during the initial fiddling
> with this. If we leave it const, we get:
>
> CC arch/arm64/kernel/asm-offsets.s
> In file included from ./include/linux/socket.h:8,
> from ./include/linux/compat.h:15,
> from ./arch/arm64/include/asm/ftrace.h:52,
> from ./include/linux/ftrace.h:23,
> from arch/arm64/kernel/asm-offsets.c:12:
> ./include/linux/uio.h: In function ‘iov_iter_ubuf’:
> ./include/linux/uio.h:374:12: error: assignment of read-only location ‘*i’
> 374 | *i = (struct iov_iter) {
> | ^
> make[1]: *** [scripts/Makefile.build:114: arch/arm64/kernel/asm-offsets.s] Error 1
> make: *** [Makefile:1286: prepare0] Error 2
>
> Let me take a closer look at that...
We can get rid of these if we convert the iov_iter initializers to
just assign the members rather than the copy+zero fill. The automatic
zero fill is nice though, in terms of sanity.
--
Jens Axboe
