Andrew Morton <[EMAIL PROTECTED]> writes: >> On Fri, 9 Mar 2007 01:13:14 +0100 "Rafael J. Wysocki" <[EMAIL PROTECTED]> >> wrote: >> I get the following traces from 2.6.21-rc3-mm2 during the "resume" phase >> of testing with 'echo test > /sys/power/disk && echo disk > >> /sys/power/state': >> >> acpi thermal:00: resuming >> pci 0000:00:00.0: resuming >> pcieport-driver 0000:00:01.0: resuming >> BUG: at drivers/pci/pci.c:679 pci_restore_state() >> >> Call Trace: >> [<ffffffff803ab3e9>] pci_restore_state+0x229/0x270 >> [<ffffffff803b10f9>] pcie_portdrv_restore_config+0x19/0x40 >> [<ffffffff803b1171>] pcie_portdrv_resume+0x11/0x20 >> [<ffffffff803ad92c>] pci_device_resume+0x2c/0x70 >> [<ffffffff80411d41>] resume_device+0xe1/0x160 >> [<ffffffff80411e69>] dpm_resume+0xa9/0x110 >> [<ffffffff80411f18>] device_resume+0x48/0x60 >> [<ffffffff802bb355>] pm_suspend_disk+0x235/0x250 >> [<ffffffff802b99c5>] enter_state+0x65/0x250 >> [<ffffffff802b9c2a>] state_store+0x7a/0xa0 >> [<ffffffff80316734>] subsys_attr_store+0x24/0x30 >> [<ffffffff803168a0>] sysfs_write_file+0x100/0x140 >> [<ffffffff802181cf>] vfs_write+0xdf/0x180 >> [<ffffffff80218d10>] sys_write+0x50/0x90 >> [<ffffffff8026529e>] system_call+0x7e/0x83 > > Yes, a number of people (including myself) have been hitting these > new warnings. I don't think we know why yet, but Eric is offline > for a bit and I'm travelling. We'll sort it out over the next few > weeks I guess.
I'm online again. I had fun getting caught in the trailing edge of a blizzard in Nebraska earlier but I have found my way back to my apartment and my computer and friends. Jeff Garzik is the one who really spotted what the problem is. He pointed out that pci_save_state and pci_restore_state have not historically required being paired (as the usually are during suspend and resume) and there is a practical use in resetting devices for not requiring them to be paired. The recent additions to pci_save_state and pci_restore_state for the msi, pci-e, pci-x all assumed those calls would be paired, and thus allocated a buffer on save and freed the buffer and restore. My WARN_ON's tested to ensure all of the buffers were freed. However drivers like the tg3 that use pci_save/restore_state for more than just suspend/resume wind up triggering the warning. I have recently sent out two patches to remove the pairing requirement but I haven't figured out suspend/resume for any of my machines so I couldn't test them. Eric - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/