Re: [dpdk-dev] Printfs in Signal Handler

[Thomas Monjalon <thomas@monjalon.net>]

04/12/2020 10:20, prateekag:
> Hi,
> 
> I can only look at this problem a week later, If someone wants to correct 
> this, they may go ahead. The problem is how to convert an integer to string 
> and what functions to use to create the final string. "itoa" is not a part of 
> C standard, writing custom itoa may pollute the code. strcpy is asyc-safe 
> afaik.
> The strings in which signal number is not required to printed can be directly 
> converted to write call.

First question is to decide whether the prints are useful enough
to be kept?


> On 2020-12-03 02:11, Thomas Monjalon wrote:
> > 01/12/2020 10:24, Thomas Monjalon:
> >> 01/12/2020 04:39, prateekag:
> >> > printf is not signal safe and may lead to deadlock if kept in signal handler
> >> > and
> >> > signal comes in two quick succession. It will not lead to incorrect behavior
> >> > and it is a highly unlikely event. And an expert may understand why this
> >> > issue happened. Is it worthwhile to change these printfs to unix I/O write
> >> > call?
> >> 
> >> Thank you for reporting.
> >> Please could you describe which files and functions you are talking about?
> > 
> > I've received a private reply:
> > 	Filenames:
> > 		testpmd.c
> > 		test_pmd_perf.c
> > 		evt_main.c
> > 		pdump/main.c
> > 	Function name:
> > 		signal_handler
> > 
> > 
> > I am OK with removing these printfs.
> > 
> > Do you want to do the patch with the explanation?
>