On Fri, May 08, 2020 at 04:36:28PM +0200, Mark Kettenis wrote:
> > Date: Fri, 8 May 2020 15:39:26 +0200
> > From: Jan Klemkow
> > On Mon, Apr 27, 2020 at 10:42:52AM +0200, Martin Pieuchot wrote:
> > > So the current logic assumes that as long as the ring contains descriptors
> > > em_rxrefill() fa
> Date: Sat, 9 May 2020 12:07:55 +0200
> From: Jan Klemkow
>
> On Fri, May 08, 2020 at 04:36:28PM +0200, Mark Kettenis wrote:
> > > Date: Fri, 8 May 2020 15:39:26 +0200
> > > From: Jan Klemkow
> > > On Mon, Apr 27, 2020 at 10:42:52AM +0200, Martin Pieuchot wrote:
> > > > So the current logic ass
On Mon, Apr 27, 2020 at 10:42:52AM +0200, Martin Pieuchot wrote:
> On 21/04/20(Tue) 15:54, Jan Klemkow wrote:
> > The following diff fixes a deadlock in em(4) when system runs out of
> > mbuf(9)s.
> >
> > Tested with current on amd64 with:
> > em0 at pci0 dev 25 function 0 "Intel I217-LM" rev 0x05
> Date: Fri, 8 May 2020 15:39:26 +0200
> From: Jan Klemkow
>
> On Mon, Apr 27, 2020 at 10:42:52AM +0200, Martin Pieuchot wrote:
> >
> > So the current logic assumes that as long as the ring contains descriptors
> > em_rxrefill() failures aren't fatal and it is not necessary to schedule a
> > ref
On 21/04/20(Tue) 15:54, Jan Klemkow wrote:
> Hi,
>
> The following diff fixes a deadlock in em(4) when system runs out of
> mbuf(9)s.
>
> Tested with current on amd64 with:
> em0 at pci0 dev 25 function 0 "Intel I217-LM" rev 0x05: msi, mac 0x1e phy
> 0xb, address d0:50:99:c1:67:94
>
> When the
Hi,
The following diff fixes a deadlock in em(4) when system runs out of
mbuf(9)s.
Tested with current on amd64 with:
em0 at pci0 dev 25 function 0 "Intel I217-LM" rev 0x05: msi, mac 0x1e phy 0xb,
address d0:50:99:c1:67:94
When the system runs out of mbufs, the receive ring of the em-NIC also
r