Re: [dpdk-dev] [PATCH] Fix VMware emulated e1000 doesn't support INTX_DISABLE flag

["Kirill A. Korinsky" <kirill@korins.ky>]

> On 28 Sep 2015, at 01:41, Stephen Hemminger <stephen@networkplumber.org> wrote:
> 
> On Sun, 27 Sep 2015 12:51:06 +0400
> "Kirill A. Korinsky" <kirill@korins.ky> wrote:
> 
>> When you try to run any DPDK-app in VMWare with e1000/82545EM NIC
>> you have a lot of lines like:
>> 
>>  EAL: Error reading from file descriptor 9: Input/output error
>> 
>> The reason of this error is bad INTX emulation in VMWare
>> and pci_intx_mask_supported return false every time.
>> ---
>> lib/librte_eal/linuxapp/igb_uio/igb_uio.c | 4 +++-
>> 1 file changed, 3 insertions(+), 1 deletion(-)
>> 
>> diff --git a/lib/librte_eal/linuxapp/igb_uio/igb_uio.c b/lib/librte_eal/linuxapp/igb_uio/igb_uio.c
>> index 865a276..c85f7e5 100644
>> --- a/lib/librte_eal/linuxapp/igb_uio/igb_uio.c
>> +++ b/lib/librte_eal/linuxapp/igb_uio/igb_uio.c
>> @@ -36,6 +36,7 @@
>> #ifdef CONFIG_XEN_DOM0
>> #include <xen/xen.h>
>> #endif
>> +#include <asm/hypervisor.h>
>> #include <rte_pci_dev_features.h>
>> 
>> #include "compat.h"
>> @@ -508,7 +509,8 @@ igbuio_pci_probe(struct pci_dev *dev, const struct pci_device_id *id)
>> 		}
>> 		/* fall back to INTX */
>> 	case RTE_INTR_MODE_LEGACY:
>> -		if (pci_intx_mask_supported(dev)) {
>> +		/*  VMware emulated e1000 doesn't support INTX_DISABLE flag */
>> +		if (pci_intx_mask_supported(dev) || x86_hyper == &x86_hyper_vmware) {
>> 			dev_dbg(&dev->dev, "using INTX");
>> 			udev->info.irq_flags = IRQF_SHARED;
>> 			udev->info.irq = dev->irq;
> 
> I don't think driver should be expecting INTx to work on VMware.
> If kernel test fails, it means masking is broken (in VMware).
> 
> 
> It is better for driver (E1000 PMD) to report that Link state interrupt does
> not work (if IRQ setup fails). Then application can test if LSC irq works
> and not enable it.

I found temporary workaround for VMware: you must change ethernet[ID].virtualDev = "e1000" to "e1000e" in vmx configuration file this virtual machine.