Yo, > So I think we should implement the 82574l behavior?
Well, as I understand it - its already implemented. I've added ICR clearance if ICR & IMS(also need to add ICR_ASSERTED check, my bad, I'll prepare new patch). At first, I had hacks to clear 'msi_causes_pending' at 'e1000e_core_set_link_status()' before link down. It works but it's not a solution. Also, on Windows the bug doesn't reproduce. I've traced Windows and Linux - the difference that Windows driver clears pending by writing to ICR, where Linux tries to clear by reading it. I had another possible fix - for Linux driver(writing to ICR at interrupt routine). I've asked intel guys, does Linux driver works with a device(I don't have real one). Thay said that it works and suggested to check 8257x spec. I'll forward the message to you. On Sat, May 9, 2020 at 9:02 AM Jason Wang <jasow...@redhat.com> wrote: > > On 2020/5/9 上午2:13, Andrew Melnichenko wrote: > > Yo, I've used OpenSDM_8257x-18.pdf specification. > > This document was recommended by Intel guys(Also, they referenced to > > that note). > > I've made a fast fix and it works. Before that I had a fix for Linux > > e1000e driver. > > Overall, the issue was in pending interrupts that can't be cleared by > > reading ICR in Linux(Windows driver clears by writing to ICR). > > > > You can download spec for example from: > > > http://iweb.dl.sourceforge.net/project/e1000/8257x%20Developer%20Manual/Revision%201.8/OpenSDM_8257x-18.pdf > > > Interesting, this spec doesn't include 82574l which is what e1000e > claims to emulate: > > c->vendor_id = PCI_VENDOR_ID_INTEL; > c->device_id = E1000_DEV_ID_82574L; > > Looking at 82574l spec (using the link mentioned in e1000e_core.c), it > said (7.4.3): > > In MSI-X mode the bits in this register can be configured to auto-clear > when the MSI-X > interrupt message is sent, in order to minimize driver overhead, and > when using MSI-X > interrupt signaling. > In systems that do not support MSI-X, reading the ICR register clears > it's bits or writing > 1b's clears the corresponding bits in this register. > > So the auto clear is under the control of EIAC (MSIX) or unconditionally > (non MSI-X). > > But what has been implemented in e1000e_mac_icr_read() is something > similar to the behavior of non 82574l card. > > So I think we should implement the 82574l behavior? > > Thanks > > > > > > On Fri, May 8, 2020 at 5:21 AM Jason Wang <jasow...@redhat.com > > <mailto:jasow...@redhat.com>> wrote: > > > > > > On 2020/5/7 上午5:26, and...@daynix.com <mailto:and...@daynix.com> > > wrote: > > > From: Andrew Melnychenko <and...@daynix.com > > <mailto:and...@daynix.com>> > > > > > > Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1707441 > > > Added ICR clearing if there is IMS bit - according to the note by > > > section 13.3.27 of the 8257X developers manual. > > > > > > Signed-off-by: Andrew Melnychenko <and...@daynix.com > > <mailto:and...@daynix.com>> > > > --- > > > hw/net/e1000e_core.c | 9 +++++++++ > > > hw/net/trace-events | 1 + > > > 2 files changed, 10 insertions(+) > > > > > > diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c > > > index d5676871fa..302e99ff46 100644 > > > --- a/hw/net/e1000e_core.c > > > +++ b/hw/net/e1000e_core.c > > > @@ -2624,6 +2624,15 @@ e1000e_mac_icr_read(E1000ECore *core, int > > index) > > > e1000e_clear_ims_bits(core, core->mac[IAM]); > > > } > > > > > > + /* > > > + * PCIe* GbE Controllers Open Source Software Developer's > > Manual > > > + * 13.3.27 Interrupt Cause Read Register > > > + */ > > > > > > Hi Andrew: > > > > Which version of the manual did you use? I try to use the one > > mentioned > > in e1000e.c which is > > > http://www.intel.com/content/dam/doc/datasheet/82574l-gbe-controller-datasheet.pdf > . > > > > But I couldn't find chapter 13.3.27. > > > > Thanks > > > > > > > + if (core->mac[ICR] & core->mac[IMS]) { > > > + trace_e1000e_irq_icr_clear_icr_bit_ims(core->mac[ICR], > > core->mac[IMS]); > > > + core->mac[ICR] = 0; > > > + } > > > + > > > trace_e1000e_irq_icr_read_exit(core->mac[ICR]); > > > e1000e_update_interrupt_state(core); > > > return ret; > > > diff --git a/hw/net/trace-events b/hw/net/trace-events > > > index e18f883cfd..46e40fcfa9 100644 > > > --- a/hw/net/trace-events > > > +++ b/hw/net/trace-events > > > @@ -237,6 +237,7 @@ e1000e_irq_icr_read_entry(uint32_t icr) > > "Starting ICR read. Current ICR: 0x%x" > > > e1000e_irq_icr_read_exit(uint32_t icr) "Ending ICR read. > > Current ICR: 0x%x" > > > e1000e_irq_icr_clear_zero_ims(void) "Clearing ICR on read due > > to zero IMS" > > > e1000e_irq_icr_clear_iame(void) "Clearing ICR on read due to > IAME" > > > +e1000e_irq_icr_clear_icr_bit_ims(uint32_t icr, uint32_t ims) > > "Clearing ICR on read due corresponding IMS bit: 0x%x & 0x%x" > > > e1000e_irq_iam_clear_eiame(uint32_t iam, uint32_t cause) > > "Clearing IMS due to EIAME, IAM: 0x%X, cause: 0x%X" > > > e1000e_irq_icr_clear_eiac(uint32_t icr, uint32_t eiac) > > "Clearing ICR bits due to EIAC, ICR: 0x%X, EIAC: 0x%X" > > > e1000e_irq_ims_clear_set_imc(uint32_t val) "Clearing IMS bits > > due to IMC write 0x%x" > > > >