On 1/7/14, Guillem Jover <[email protected]> wrote:
> I think adding OSDragonflyBSD to previous preprocessor block for the
> other BSD should be enough, the next one catching HAVE_KVM_H should
> include the remaning header files.
It's a problematic because dfly disallows including sys/proc.h from userspace.
>> +#elif defined (OSDragonflyBSD)
>> +#define PROCESS_NAME_SIZE MAXCOMLEN
>> #endif
>
> Just to know what's the portably safe minimum for callers to use, to
> what does MAXCOMLEN expand?
16. But I think it's better to use system defined macro then "magic"
constant. FreeBSD has MAXCOMLEN too.
> I've got a patch to switch the FreeBSD code to use the KVM method,
> unfortunately it seems to use ki_ namespaced struct members.
>
> Also the KVM method seems to be missing a proper do_procinit(), which
> could be something like the following untested code, I had lying
> around:
>
> ,---
> static enum status_code
> do_procinit(void)
> {
> - /* Nothing to do. */
> - return status_unknown;
> + kvm_t *kd;
> + int nentries, i;
> + struct kinfo_proc *kp;
> + char errbuf[_POSIX2_LINE_MAX];
> + enum status_code prog_status = status_dead;
> +
> + kd = kvm_openfiles(NULL, NULL, NULL, O_RDONLY, errbuf);
> + if (kd == NULL)
> + errx(1, "%s", errbuf);
> + kp = kvm_getprocs(kd, KERN_PROC_PROC, 0, &nentries);
> + if (kp == NULL)
> + errx(1, "%s", kvm_geterr(kd));
> +
> + for (i = 0; i < nentries; i++) {
> + enum status_code pid_status;
> +
> + pid_status = pid_check(kp[i].ki_pid);
> + if (pid_status < prog_status)
> + prog_status = pid_status;
> + }
> +
> + kvm_close(kd);
> +
> + return prog_status;
> }
> `---
>
> For DragonFly BSD, you'll need to change ki_pid to kp_pid I guess. If
> this works on DragonFly BSD, I'll commit that before applying your patch.
It works, but with some more dfly specific tweaks (it uses
KERN_PROC_ALL where other OSes uses KERN_PROC_PROC, and its kernel
PIDs are less than 1).
I'll send refreshed dfly patch after you apply patch for do_procinit.
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
