Re: [PATCH v4 RFC liburing 3/3] Wrap all syscalls in a kernel style return value

[Ammar Faizi <ammar.faizi@students.amikom.ac.id>]

On Sun, Oct 3, 2021 at 8:19 PM Jens Axboe <axboe@kernel.dk> wrote:
>> diff --git a/src/register.c b/src/register.c
>> index cb09dea..fec144d 100644
>> --- a/src/register.c
>> +++ b/src/register.c
>> @@ -6,7 +6,6 @@
>>  #include <sys/mman.h>
>>  #include <sys/resource.h>
>>  #include <unistd.h>
>> -#include <errno.h>
>>  #include <string.h>
>>  
>>  #include "liburing/compat.h"
>> @@ -104,13 +103,16 @@ int io_uring_register_files_update(struct io_uring *ring, unsigned off,
>>  
>>  static int increase_rlimit_nofile(unsigned nr)
>>  {
>> +	int ret;
>>  	struct rlimit rlim;
>>  
>> -	if (getrlimit(RLIMIT_NOFILE, &rlim) < 0)
>> -		return -errno;
>> +	ret = uring_getrlimit(RLIMIT_NOFILE, &rlim);
>> +	if (ret < 0)
>> +		return ret;
>> +
>>  	if (rlim.rlim_cur < nr) {
>>  		rlim.rlim_cur += nr;
>> -		setrlimit(RLIMIT_NOFILE, &rlim);
>> +		return uring_setrlimit(RLIMIT_NOFILE, &rlim);
>>  	}
>
>This isn't a functionally equivalent transformation, and it's
>purposefully not returning failure to increase. It may still succeed if
>we fail here, relying on failure later for the actual operation that
>needs an increase in files.
>
>> diff --git a/src/syscall.h b/src/syscall.h
>> index f7f63aa..3e964ed 100644
>> --- a/src/syscall.h
>> +++ b/src/syscall.h
>> @@ -4,11 +4,15 @@
>>  
>>  #include <errno.h>
>>  #include <signal.h>
>> +#include <stdint.h>
>>  #include <unistd.h>
>> +#include <stdbool.h>
>>  #include <sys/mman.h>
>>  #include <sys/syscall.h>
>>  #include <sys/resource.h>
>>  
>> +#include <liburing.h>
>> +
>>  #ifdef __alpha__
>>  /*
>>   * alpha and mips are exception, other architectures have
>> @@ -60,6 +64,21 @@ int __sys_io_uring_register(int fd, unsigned int opcode, const void *arg,
>>  			    unsigned int nr_args);
>>  
>>  
>> +static inline void *ERR_PTR(intptr_t n)
>> +{
>> +	return (void *) n;
>> +}
>> +
>> +
>
>Extra newline here.
>
>Apart from those two, starting to look pretty reasonable.
>
>-- 
>Jens Axboe
>

Thanks for the review, I will address those two and send the v5 of
this series.

-- 
Ammar Faizi