[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] [PATCH] esx: Disable libcurl's use of signals to fix a segfault

2012/10/1 Michal Privoznik <mprivozn redhat com>:
> On 29.09.2012 22:17, Matthias Bolte wrote:
>> libcurl uses a SIGALRM in combination with sigsetjmp/siglongjmp to be
>> able to abort a DNS lookup when it takes too long. The problem with this
>> in a multi-threaded application is that the signal handler for SIGALRM
>> and the call to siglongjmp can be executed on a thread that is different
>> from the one that initially did the SIGALRM setup and the call to
>> sigsetjmp. In the reported case this triggered a segfault.
>> Disable libcurl's use of signals to avoid this situation. This has the
>> disadvantage of losing the ability to abort DNS lookups which might result
>> in libcurl getting stuck in a DNS lookup in the worst case.
>> Reported by Benjamin Wang.
>> ---
>>  src/esx/esx_vi.c |    1 +
>>  1 files changed, 1 insertions(+), 0 deletions(-)
>> diff --git a/src/esx/esx_vi.c b/src/esx/esx_vi.c
>> index 275b858..125eaee 100644
>> --- a/src/esx/esx_vi.c
>> +++ b/src/esx/esx_vi.c
>> @@ -330,6 +330,7 @@ esxVI_CURL_Connect(esxVI_CURL *curl, esxUtil_ParsedUri *parsedUri)
>>      }
>>      curl_easy_setopt(curl->handle, CURLOPT_USERAGENT, "libvirt-esx");
>> +    curl_easy_setopt(curl->handle, CURLOPT_NOSIGNAL, 1);
>>      curl_easy_setopt(curl->handle, CURLOPT_HEADER, 0);
>>      curl_easy_setopt(curl->handle, CURLOPT_FOLLOWLOCATION, 0);
>>      curl_easy_setopt(curl->handle, CURLOPT_SSL_VERIFYPEER,
> I think this is safe; From the version 7.10 curl can still timeout on
> DNS lookup even with CURLOPT_NOSIGNAL set:
>    http://curl.haxx.se/libcurl/c/curl_easy_setopt.html#CURLOPTNOSIGNAL
> So ACK.
> Michal

libcurl supports different DNS backends. This problem only affects the
synchronous one where libcurl uses the described timeout mechanism.
The asynchronous c-ares backend is not affected.

But as I don't see a better solution to avoid the thread-unsafety of
the SIGALRM/sigsetjmp/siglongjmp combination I'll push thins one.

Matthias Bolte

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]