| Message ID | b8e7df14-d95e-4aab-b0e3-3b90ae0d3c21@gmail.com (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] r8169: avoid unsolicited interrupts | expand |
On Thu, Oct 17, 2024 at 08:05:16AM +0200, Heiner Kallweit wrote: > It was reported that after resume from suspend a PCI error is logged > and connectivity is broken. Error message is: > PCI error (cmd = 0x0407, status_errs = 0x0000) > The message seems to be a red herring as none of the error bits is set, > and the PCI command register value also is normal. Exception handling > for a PCI error includes a chip reset what apparently brakes connectivity > here. The interrupt status bit triggering the PCI error handling isn't > actually used on PCIe chip versions, so it's not clear why this bit is > set by the chip. > Fix this by ignoring interrupt status bits which aren't part of the > interrupt mask. > Note that RxFIFOOver needs a special handling on certain chip versions, > it's handling isn't changed with this patch. > > Fixes: 0e4851502f84 ("r8169: merge with version 8.001.00 of Realtek's r8168 driver") > Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219388 > Tested-by: Pengyu Ma <mapengyu@gmail.com> > Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Reviewed-by: Simon Horman <horms@kernel.org>
On 17.10.2024 08:05, Heiner Kallweit wrote: > It was reported that after resume from suspend a PCI error is logged > and connectivity is broken. Error message is: > PCI error (cmd = 0x0407, status_errs = 0x0000) > The message seems to be a red herring as none of the error bits is set, > and the PCI command register value also is normal. Exception handling > for a PCI error includes a chip reset what apparently brakes connectivity > here. The interrupt status bit triggering the PCI error handling isn't > actually used on PCIe chip versions, so it's not clear why this bit is > set by the chip. > Fix this by ignoring interrupt status bits which aren't part of the > interrupt mask. > Note that RxFIFOOver needs a special handling on certain chip versions, > it's handling isn't changed with this patch. > > Fixes: 0e4851502f84 ("r8169: merge with version 8.001.00 of Realtek's r8168 driver") > Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219388 > Tested-by: Pengyu Ma <mapengyu@gmail.com> > Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> > --- When doing some unrelated performance tests I found that this patch breaks connectivity under heavy load. Please drop it. I'll investigate and come up with an alternative way to fix the reported issue.
On Thu, Oct 17, 2024 at 11:22:20PM +0200, Heiner Kallweit wrote: > On 17.10.2024 08:05, Heiner Kallweit wrote: > > It was reported that after resume from suspend a PCI error is logged > > and connectivity is broken. Error message is: > > PCI error (cmd = 0x0407, status_errs = 0x0000) > > The message seems to be a red herring as none of the error bits is set, > > and the PCI command register value also is normal. Exception handling > > for a PCI error includes a chip reset what apparently brakes connectivity > > here. The interrupt status bit triggering the PCI error handling isn't > > actually used on PCIe chip versions, so it's not clear why this bit is > > set by the chip. > > Fix this by ignoring interrupt status bits which aren't part of the > > interrupt mask. > > Note that RxFIFOOver needs a special handling on certain chip versions, > > it's handling isn't changed with this patch. > > > > Fixes: 0e4851502f84 ("r8169: merge with version 8.001.00 of Realtek's r8168 driver") > > Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219388 > > Tested-by: Pengyu Ma <mapengyu@gmail.com> > > Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> > > --- > > When doing some unrelated performance tests I found that this patch breaks > connectivity under heavy load. Please drop it. > I'll investigate and come up with an alternative way to fix the reported issue. You should be able to mark your own patches are change request etc. Andrew --- pw-bot: cr
On 18.10.2024 04:52, Andrew Lunn wrote: > On Thu, Oct 17, 2024 at 11:22:20PM +0200, Heiner Kallweit wrote: >> On 17.10.2024 08:05, Heiner Kallweit wrote: >>> It was reported that after resume from suspend a PCI error is logged >>> and connectivity is broken. Error message is: >>> PCI error (cmd = 0x0407, status_errs = 0x0000) >>> The message seems to be a red herring as none of the error bits is set, >>> and the PCI command register value also is normal. Exception handling >>> for a PCI error includes a chip reset what apparently brakes connectivity >>> here. The interrupt status bit triggering the PCI error handling isn't >>> actually used on PCIe chip versions, so it's not clear why this bit is >>> set by the chip. >>> Fix this by ignoring interrupt status bits which aren't part of the >>> interrupt mask. >>> Note that RxFIFOOver needs a special handling on certain chip versions, >>> it's handling isn't changed with this patch. >>> >>> Fixes: 0e4851502f84 ("r8169: merge with version 8.001.00 of Realtek's r8168 driver") >>> Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219388 >>> Tested-by: Pengyu Ma <mapengyu@gmail.com> >>> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> >>> --- >> >> When doing some unrelated performance tests I found that this patch breaks >> connectivity under heavy load. Please drop it. >> I'll investigate and come up with an alternative way to fix the reported issue. > > You should be able to mark your own patches are change request etc. > When doing this years ago, don't recall which subsystem it was, the maintainer wasn't too happy that somebody else was changing the status of patches in patchwork. Next time I'll do so, thanks for the hint. > Andrew > Heiner > --- > pw-bot: cr
> > You should be able to mark your own patches are change request etc. > > > When doing this years ago, don't recall which subsystem it was, the maintainer > wasn't too happy that somebody else was changing the status of patches in patchwork. > Next time I'll do so, thanks for the hint. To clarify, you can change it using the pw-bot. Put pw-bot: cr somewhere in the email to set it to Change Request. https://www.kernel.org/doc/html/latest/process/maintainer-netdev.html#updating-patch-status Andrew
diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c index 1a4d834c2..322a1e930 100644 --- a/drivers/net/ethernet/realtek/r8169_main.c +++ b/drivers/net/ethernet/realtek/r8169_main.c @@ -4805,7 +4805,11 @@ static irqreturn_t rtl8169_interrupt(int irq, void *dev_instance) struct rtl8169_private *tp = dev_instance; u32 status = rtl_get_events(tp); - if ((status & 0xffff) == 0xffff || !(status & tp->irq_mask)) + if ((status & 0xffff) == 0xffff) + return IRQ_NONE; + + status &= tp->irq_mask | RxFIFOOver; + if (!status) return IRQ_NONE; if (unlikely(status & SYSErr)) {
It was reported that after resume from suspend a PCI error is logged and connectivity is broken. Error message is: PCI error (cmd = 0x0407, status_errs = 0x0000) The message seems to be a red herring as none of the error bits is set, and the PCI command register value also is normal. Exception handling for a PCI error includes a chip reset what apparently brakes connectivity here. The interrupt status bit triggering the PCI error handling isn't actually used on PCIe chip versions, so it's not clear why this bit is set by the chip. Fix this by ignoring interrupt status bits which aren't part of the interrupt mask. Note that RxFIFOOver needs a special handling on certain chip versions, it's handling isn't changed with this patch. Fixes: 0e4851502f84 ("r8169: merge with version 8.001.00 of Realtek's r8168 driver") Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219388 Tested-by: Pengyu Ma <mapengyu@gmail.com> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> --- drivers/net/ethernet/realtek/r8169_main.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)