[CRIU] [PATCH 06/12] crtools: dump pending signals

[Andrey Wagin]

2012/12/18 Pavel Emelyanov <xemul at parallels.com>:
>> @@ -941,6 +942,12 @@ static int dump_task_core_all(pid_t pid, const struct proc_pid_stat *stat,
>>       if (ret < 0)
>>               goto err_free;
>>
>> +     fd_core = open_image(CR_FD_PSIGNAL, O_DUMP, item->pid.virt);
>> +     if (fd_core < 0)
>> +             goto err_free;
>> +     ret = parasite_dump_signal_seized(pid, ctl, fd_core, 0);
>> +     close(fd_core);
>> +
>>  err_free:
>>       core_entry_free(core);
>>       pr_info("----------------------------------------\n");
>> @@ -1567,6 +1584,13 @@ static int dump_one_task(struct pstree_item *item)
>>               goto err_cure;
>>       }
>>
>> +     ret = parasite_dump_signal_seized(pid, parasite_ctl,
>> +                                     fdset_fd(cr_fdset, CR_FD_SIGNAL), 1);
>> +     if (ret) {
>> +             pr_err("Can't dump pending signals (pid: %d)\n", pid);
>> +             goto err_cure;
>> +     }
>> +
>>       ret = parasite_dump_sigacts_seized(parasite_ctl, cr_fdset);
>>       if (ret) {
>>               pr_err("Can't dump sigactions (pid: %d) with parasite\n", pid);
>
> Why can't we go to parasite code once? Or twice, but in one place?
>

Here is another problem. A thread args for a thread leader is saved
separate from other thread_args-s. I'm going to fix that.