We don't want the DSP to continue writing into other mapped pages, no matter how unlikely. Based on extensive discussion with Fernando Guzman Lugo. Signed-off-by: Felipe Contreras <felipe.contreras@xxxxxxxxx> --- drivers/staging/tidspbridge/core/ue_deh.c | 8 ++++++++ 1 files changed, 8 insertions(+), 0 deletions(-) diff --git a/drivers/staging/tidspbridge/core/ue_deh.c b/drivers/staging/tidspbridge/core/ue_deh.c index 06167ed..793e982 100644 --- a/drivers/staging/tidspbridge/core/ue_deh.c +++ b/drivers/staging/tidspbridge/core/ue_deh.c @@ -186,6 +186,14 @@ void bridge_deh_notify(struct deh_mgr *deh_mgr, u32 ulEventMask, u32 dwErrInfo) print_dsp_trace_buffer(dev_context); dump_dl_modules(dev_context); + /* + * Before acking the MMU fault, let's make sure MMU can only + * access entry #0. Then add a new entry so that the DSP OS + * can continue in order to dump the stack. + */ + hw_mmu_twl_disable(resources->dw_dmmu_base); + hw_mmu_tlb_flush_all(resources->dw_dmmu_base); + hw_mmu_tlb_add(resources->dw_dmmu_base, virt_to_phys(dummy_va_addr), fault_addr, HW_PAGE_SIZE4KB, 1, -- 1.7.1 -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
