Gilles Chanteperdrix wrote: > Jan Kiszka wrote: >> Gilles Chanteperdrix wrote: >>> Jan Kiszka wrote: >>>> Gilles Chanteperdrix wrote: >>>>> Jan Kiszka wrote: >>>>>> Gilles Chanteperdrix wrote: >>>>>>> Jan Kiszka wrote: >>>>>>>> BUG_ON(pgd_page_vaddr(*pgd) != >>>>>>>> pgd_page_vaddr(*pgd_ref)); >>>>> Ok. Maybe printing pgd_page_vaddr(*pgd) and pgd_page_vaddr(*pgd_ref) >>>>> would help ? >>>> I don't see yet where you want to go. As I said, the issue is rare. I >>>> rather think we need to approach it theoretically. >>> Well the bug happens when the two virtual addresses are different (so, I >>> do not think my original remark applies). So, I want to see what the two >>> addresses look like, for instance if they are not 0 or ~0. >> OK, will forward the request. > > Ok. Something else: x86_32 and vmalloc_sync_all use spinlock_irqsave, > so, maybe __ipipe_pin_range_globally should use irqsave as well ?
Hmm. In 2.6.24 this lock was not used with irqsave, but in 2.6.26 it is. One would have to check if this is a bug fix or due to other changes. Jan -- Siemens AG, Corporate Technology, CT SE 2 Corporate Competence Center Embedded Linux _______________________________________________ Adeos-main mailing list [email protected] https://mail.gna.org/listinfo/adeos-main
