On 02/04/2014 03:29 PM, John Ferlan wrote:
On 02/04/2014 06:06 AM, Ján Tomko wrote:
> On 01/30/2014 06:50 PM, John Ferlan wrote:
>> + VIR_FREE(errbuf);
>> + goto cleanup;
>> }
>>
>> goto recheck;
>> }
>>
>> + /* If we know failure was because of blacklist, let's report that */
>> + if (virKModIsBlacklisted(driver)) {
>> + virReportError(VIR_ERR_INTERNAL_ERROR,
>> + _("Failed to load PCI stub module %s: "
>> + "administratively prohibited"),
>> + driver);
>> + }
>> +
>> +cleanup:
>> return -1;
>> }
>>
>> @@ -1313,9 +1322,10 @@ virPCIDeviceDetach(virPCIDevicePtr dev,
>> virPCIDeviceList *inactiveDevs)
>> {
>> if (virPCIProbeStubDriver(dev->stubDriver) < 0) {
>> - virReportError(VIR_ERR_INTERNAL_ERROR,
>> - _("Failed to load PCI stub module %s"),
>> - dev->stubDriver);
>> + if (virGetLastError() == NULL)
>
> This seems to be the only caller of virPCIProbeStubDriver.
> You could just report the error unconditionally inside it.
>
Attempting to make the differentiation between load failed load failed
because of administratively prohibited means an additional check or two
in the caller.
I meant that right now virPCIProbeStubDriver is only called from here and if
it did not report an error, we will report one here.
It seemed cleaner not to report an error here and make virPCIProbeStubDriver
report an error in all cases (not just when the module is blacklisted and/or
on OOM in virPCIDriverDir).
Furthermore if something that virPCIProbeStubDriver() called provided
some other error wouldn't it be better to not overwrite the message? If
the virAsprintf() called by virPCIDriverDir() failed because of memory
allocation, then which error message would be displayed without the
virGetLastError() check? I guess I'm not 100% clear in my mind which
error message would get displayed...
Only the last reported error gets displayed, but both will get logged.
Jan