netdev - Re: [RFC iproute2-next] devlink: display elapsed time during flash update

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <a4250983-4520-1390-3a7c-b6bc794f04eb@pensando.io>
Date:   Tue, 29 Sep 2020 11:13:33 -0700
From:   Shannon Nelson <snelson@...sando.io>
To:     Jacob Keller <jacob.e.keller@...el.com>,
        Jakub Kicinski <kubakici@...pl>
Cc:     netdev@...r.kernel.org
Subject: Re: [RFC iproute2-next] devlink: display elapsed time during flash
 update

On 9/29/20 10:56 AM, Jacob Keller wrote:
> On 9/29/2020 10:18 AM, Jakub Kicinski wrote:
>> On Mon, 28 Sep 2020 16:49:45 -0700 Jacob Keller wrote:
>>> For some devices, updating the flash can take significant time during
>>> operations where no status can meaningfully be reported. This can be
>>> somewhat confusing to a user who sees devlink appear to hang on the
>>> terminal waiting for the device to update.
>>>
>>> Provide a ticking counter of the time elapsed since the previous status
>>> message in order to make it clear that the program is not simply stuck.
>>>
>>> Do not display this message unless a few seconds have passed since the
>>> last status update. Additionally, if the previous status notification
>>> included a timeout, display this as part of the message. If we do not
>>> receive an error or a new status without that time out, replace it with
>>> the text "timeout reached".
>>>
>>> Signed-off-by: Jacob Keller <jacob.e.keller@...el.com>
>>> ---
>>> Sending this as an RFC because I doubt this is the best implementation. For
>>> one, I get a weird display issue where the cursor doesn't always end up on
>>> the end of line in my shell.. The % display works properly, so I'm not sure
>>> what's wrong here.
>>>
>>> Second, even though select should be timing out every 1/10th of a second for
>>> screen updates, I don't seem to get that behavior in my test. It takes about
>>> 8 to 10 seconds for the first elapsed time message to be displayed, and it
>>> updates really slowly. Is select just not that precise? I even tried using a
>>> timeout of zero, but this means we refresh way too often and it looks bad. I
>>> am not sure what is wrong here...
>> Strange. Did you strace it? Perhaps it's some form of output buffering?
>>
> Haven't yet, just noticed the weird output behavior and timing
> inconsistency.

Add a fflush(stdout) at the end of pr_out_tty()?  Or maybe simply 
disable stdout buffering with setvbuf().

>>> diff --git a/devlink/devlink.c b/devlink/devlink.c
>>> index 0374175eda3d..7fb4b5ef1ebe 100644
>>> --- a/devlink/devlink.c
>>> +++ b/devlink/devlink.c
>>> @@ -33,6 +33,7 @@
>>>   #include <sys/select.h>
>>>   #include <sys/socket.h>
>>>   #include <sys/types.h>
>>> +#include <sys/time.h>
>>>   #include <rt_names.h>
>>>   
>>>   #include "version.h"
>>> @@ -3066,6 +3067,9 @@ static int cmd_dev_info(struct dl *dl)
>>>   
>>>   struct cmd_dev_flash_status_ctx {
>>>   	struct dl *dl;
>>> +	struct timeval last_status_msg;

Having some 'time' concept in this name will make it less likely for the 
reader to confuse it with something that holds a text message - 't_', or 
'_secs' or something.

sln