Most recent kernel where this bug did not occur: N/A Distribution: any Hardware Environment: amd64 laptop nforce3 chipset ati radeon m9 videocard Software Environment: N/A Problem Description: after resume from acpi S3 suspend-to-ram text console works but system hangs when starting X (or switching to virtual console where X is running). No useful messages was captured via network console. I was able to produce workaround for this bug, here it is (this patch is for 2.6.16-rc5 kernel): drivers/char/agp/amd64-agp.c | 1 + 1 file changed, 1 insertion(+) Index: linux/drivers/char/agp/amd64-agp.c =================================================================== --- linux.orig/drivers/char/agp/amd64-agp.c +++ linux/drivers/char/agp/amd64-agp.c @@ -615,6 +615,7 @@ static int agp_amd64_resume(struct pci_d pci_set_power_state(pdev, PCI_D0); pci_restore_state(pdev); + nforce3_agp_init(pdev); return amd_8151_configure(); } with this patch, everything works just fine. Steps to reproduce: (1). compile 2.6.16-rc5 kernel, applying hack mentioned here: http://bugzilla.kernel.org/show_bug.cgi?id=5866#c1 to make hdd work after resume. (2). suspend system, saving video state with vbetool version 0.4 or later (previous versions of vbetool did not work on amd64) http://www.srcf.ucam.org/~mjg59/vbetool/ as prescribed in Documentation/power/video.txt (3). after resume, start X. use power switch to shut down hung system =). Note: this failure can be reproduced with earlier kernels, 2.6.15 and 2.6.14 (not tried 2.6.13) but the workaround for this is slightly different. (at resume, you need to call also amd_8151_configure() in amd64-agp.c and ati_configure() in ati-agp.c).
Dave, can you comment on this issue?
patch is only right on nforce machines. we don't want to be unconditionally calling that function. I'll fix it up.
This issue is already fixed in kernel 2.6.17.