Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Message-ID: <a0665d21-a390-6263-0018-09b4cb57e87b@ozlabs.ru>
Date:   Tue, 3 Aug 2021 23:13:57 +1000
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:90.0) Gecko/20100101
 Thunderbird/90.0
Subject: Re: [RFC PATCH kernel] KVM: Stop leaking memory in debugfs
Content-Language: en-US
To:     Paolo Bonzini <pbonzini@redhat.com>,
        Greg KH <gregkh@linuxfoundation.org>
Cc:     "Kernel Mailing List, Linux" <linux-kernel@vger.kernel.org>,
        kvm <kvm@vger.kernel.org>
References: <20210730043217.953384-1-aik@ozlabs.ru>
 <YQklgq4NkL4UToVY@kroah.com>
 <CABgObfb+M9Qeow1EZy+eQwM1jwoZY3zdPJfZW+Q+MoWmkaqcFw@mail.gmail.com>
From:   Alexey Kardashevskiy <aik@ozlabs.ru>
In-Reply-To: <CABgObfb+M9Qeow1EZy+eQwM1jwoZY3zdPJfZW+Q+MoWmkaqcFw@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 03/08/2021 22:52, Paolo Bonzini wrote:
> On Tue, Aug 3, 2021 at 1:16 PM Greg KH <gregkh@linuxfoundation.org> wrote:
>> On Fri, Jul 30, 2021 at 02:32:17PM +1000, Alexey Kardashevskiy wrote:
>>>        snprintf(dir_name, sizeof(dir_name), "%d-%d", task_pid_nr(current), fd);
>>>        kvm->debugfs_dentry = debugfs_create_dir(dir_name, kvm_debugfs_dir);
>>> +     if (IS_ERR_OR_NULL(kvm->debugfs_dentry)) {
>>> +             pr_err("Failed to create %s\n", dir_name);
>>> +             return 0;
>>> +     }
>>
>> It should not matter if you fail a debugfs call at all.
>>
>> If there is a larger race at work here, please fix that root cause, do
>> not paper over it by attempting to have debugfs catch the issue for you.
> 
> I don't think it's a race, it's really just a bug that is intrinsic in how
> the debugfs files are named.  You can just do something like this:
> 
> #include <unistd.h>
> #include <stdio.h>
> #include <fcntl.h>
> #include <sys/wait.h>
> #include <sys/ioctl.h>
> #include <linux/kvm.h>
> #include <stdlib.h>
> int main() {
>          int kvmfd = open("/dev/kvm", O_RDONLY);
>          int fd = ioctl(kvmfd, KVM_CREATE_VM, 0);
>          if (fork() == 0) {
>                  printf("before: %d\n", fd);
>                  sleep(2);
>          } else {
>                  close(fd);
>                  sleep(1);
>                  int fd = ioctl(kvmfd, KVM_CREATE_VM, 0);
>                  printf("after: %d\n", fd);
>                  wait(NULL);
>          }
> }

oh nice demo :)

although I still think there was a race when I saw it as there was no 
fork() in the picture but continuous create/destroy VM in 16 threads on 
16 VCPUs with no KVM_RUN in between.

> 
> So Alexey's patch is okay and I've queued it, though with pr_warn_ratelimited
> instead of pr_err.

Makes sense with your reproducer. Thanks,


-- 
Alexey