Re: [PATCH 3/4] samples/bpf: Check the error of write() and read()

[Taeung Song <treeze.taeung@gmail.com>]

Hi David Laight,

On 07/02/2018 06:25 PM, David Laight wrote:
> From: Taeung Song
>> Sent: 02 July 2018 10:15
>> test_task_rename() and test_urandom_read()
>> can be failed during write() and read(),
>> So check the result of them.
>>
>> Signed-off-by: Taeung Song <treeze.taeung@gmail.com>
>> ---
>>   samples/bpf/test_overhead_user.c | 20 +++++++++++++++-----
>>   1 file changed, 15 insertions(+), 5 deletions(-)
>>
>> diff --git a/samples/bpf/test_overhead_user.c b/samples/bpf/test_overhead_user.c
>> index 6caf47a..8a88d9c 100644
>> --- a/samples/bpf/test_overhead_user.c
>> +++ b/samples/bpf/test_overhead_user.c
>> @@ -6,6 +6,7 @@
>>    */
>>   #define _GNU_SOURCE
>>   #include <sched.h>
>> +#include <errno.h>
>>   #include <stdio.h>
>>   #include <sys/types.h>
>>   #include <asm/unistd.h>
>> @@ -44,8 +45,12 @@ static void test_task_rename(int cpu)
>>   		exit(1);
>>   	}
>>   	start_time = time_get_ns();
>> -	for (i = 0; i < MAX_CNT; i++)
>> -		write(fd, buf, sizeof(buf));
>> +	for (i = 0; i < MAX_CNT; i++) {
>> +		if (write(fd, buf, sizeof(buf)) < 0) {
>> +			printf("task rename failed: %s\n", strerror(errno));
>> +			break;
> 
> I'm not sure 'break' generates sensible output.
> 

OK,
it seems to be better to off on the above error case instead of break;

>> +		}
>> +	}
> 
> What about partial writes??
> 
Hum..
do you mean just skipping several errors of the above write() ?

>>   	printf("task_rename:%d: %lld events per sec\n",
>>   	       cpu, MAX_CNT * 1000000000ll / (time_get_ns() - start_time));
>>   	close(fd);
>> @@ -55,7 +60,7 @@ static void test_urandom_read(int cpu)
>>   {
>>   	__u64 start_time;
>>   	char buf[4];
>> -	int i, fd;
>> +	int i, fd, err = 0;
>>
>>   	fd = open("/dev/urandom", O_RDONLY);
>>   	if (fd < 0) {
>> @@ -63,8 +68,13 @@ static void test_urandom_read(int cpu)
>>   		exit(1);
>>   	}
>>   	start_time = time_get_ns();
>> -	for (i = 0; i < MAX_CNT; i++)
>> -		read(fd, buf, sizeof(buf));
>> +	for (i = 0; i < MAX_CNT; i++) {
>> +		err = read(fd, buf, sizeof(buf));
>> +		if (err < 0 || err >= sizeof(buf)) {
> 
> Overlong reads indicate that something is seriously awry.
> Short reads are valid - but maybe not expected.
>> +			printf("failed to read from /dev/urandom: %s\n", strerror(errno));
> 
> 
> strerror() won't give anything sensible unless err == -1;
> You probably want to include the loop count.
> 
>> +			break;
> 
> The summary print will be gibberish after break.
> 

Ditto, will change the code to do exit(1);
on the err == -1 case.
What do you think about it ?

Thanks,
Taeung

>> +		}
>> +	}
>>   	printf("urandom_read:%d: %lld events per sec\n",
>>   	       cpu, MAX_CNT * 1000000000ll / (time_get_ns() - start_time));
>>   	close(fd);
>> --
>> 2.7.4
>