Hi Xia, Le 01/05/2024 à 08:22, 赵夏 a écrit : > Hi Christophe, > > I identified that the function named "arch_local_irq_restore" of the > file irq_64.c was updated by you frequently in the last two years.
You mean arch/powerpc/kernel/irq_64.c I guess ? I don't think I "frequently" changed it. What I did was mainly maintenance, I first extracted all 64 bits code from irq.c into irq_64.c to ease maintenance, then I did a few cosmetic changes with no functionnal impacts. As far as I can see some more important changes were done by Nick, maybe I'll be able to answer you better. I am myself not a PPC64 expert, my domain in mainly PPC32. > In the previous code version (such as Linux Kernel 5.10.7), I thought > that an interruption such as doorbell might happen before > "local_paca->irq_happened = 0;" which can cause a problem. Is this the > correct understanding? I have two questions. > > 1. If the previous version can cause the problem, can the current fix > solve the problem? > 2. If we do not care about the os performance, can we close the > nested interrupt mechanism to avoid the problem? If we can, how to close > the nested interrupt? > > Kind regards > Xia Thanks Christophe