On 7/31/19 9:53 AM, Michal Privoznik wrote:
On 7/23/19 7:35 PM, Daniel Henrique Barboza wrote:
> This code that executes virPCIDeviceReset in all virPCIDevicePtr
> objects of a given virPCIDeviceListPtr list is replicated twice
> in the code. Putting it in a helper function helps with
> readability.
>
> Signed-off-by: Daniel Henrique Barboza <danielhb413(a)gmail.com>
> ---
> src/util/virhostdev.c | 54 +++++++++++++++++++++++--------------------
> 1 file changed, 29 insertions(+), 25 deletions(-)
>
> diff --git a/src/util/virhostdev.c b/src/util/virhostdev.c
> index a3647a6cf4..4dd24a8f65 100644
> --- a/src/util/virhostdev.c
> +++ b/src/util/virhostdev.c
> @@ -613,6 +613,32 @@
> virHostdevRestoreNetConfig(virDomainHostdevDefPtr hostdev,
> }
> }
> +static int
> +virHostdevResetAllPCIDevices(virHostdevManagerPtr mgr,
> + virPCIDeviceListPtr pcidevs)
> +{
> + int ret = 0;
> + size_t i;
> +
> + for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) {
> + virPCIDevicePtr pci = virPCIDeviceListGet(pcidevs, i);
> +
> + /* We can avoid looking up the actual device here, because
> performing
> + * a PCI reset on a device doesn't require any information
> other than
> + * the address, which 'pci' already contains */
> + VIR_DEBUG("Resetting PCI device %s", virPCIDeviceGetName(pci));
> + if (virPCIDeviceReset(pci, mgr->activePCIHostdevs,
> + mgr->inactivePCIHostdevs) < 0) {
> + VIR_ERROR(_("Failed to reset PCI device: %s"),
> + virGetLastErrorMessage());
> + virResetLastError();
I don't think this is a good idea. VIR_ERROR() logs the error message,
but then virResetLastError() clears the thread local error object and
thus if a PCI fails to reset the caller is left with 'uknown error'
and sysadmin has to go to logs to find out what's happening. In fact,
I think all those virResetLastError() calls should be removed from
virHostdevReAttachPCIDevices(). I'll fix it here and in 3/3 too.
Thanks for explaining and fixing it. I blindly moved the code to this new
function (here and in 3/3) without giving too much thought about whether
the existing error handling makes sense.
DHB
> + ret = -1;
> + }
> + }
> +
> + return ret;
> +}
> +
Michal