On Wed, 23 Jun 2021, Nikolai Zhubr wrote:

quoted

As I said before, I still think we should also merge the 8139 driver patch,
probably without that loop. It's not great, but I'm much more confident
I understand what that does and that the patched version is better than
the current code.

Yes, the 'poll' approach apparently works stable and does not cause any
measurable performance decrease. But it would need some carefull
cleanup/review, especially WRT locking. Now that all real event handling work
is happening in the poll function, it still has to be protected against the
(potentially also long-running) reset function which in current design can be
called e.g. from a different thread due to tx timeout, and this does not look
good, but it is a bit beyond my capability to arrange it better. Besides, the
idea was to keep the fix simple and avoid a massive rework...

 The simplest fix is not always the right one.

 I've now posted a small patch series that adds a PCI IRQ router for the 
SiS85C497 device.  Would you please try it with your system along with the 
debug patch included below and send me the resulting bootstap log?

  Maciej
---
 arch/x86/include/asm/pci_x86.h |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: linux-macro-ide/arch/x86/include/asm/pci_x86.h
===================================================================

--- linux-macro-ide.orig/arch/x86/include/asm/pci_x86.h
+++ linux-macro-ide/arch/x86/include/asm/pci_x86.h

@@ -7,7 +7,7 @@
 
 #include <linux/ioport.h>
 
-#undef DEBUG
+#define DEBUG 1
 
 #ifdef DEBUG
 #define DBG(fmt, ...) printk(fmt, ##__VA_ARGS__)