Hello everybody. Continuing my experiment with SL7, I tried to make more "production" system (new home router, to be exact). The problem is that system doesn't want to turn itself off with power button or "poweroff" command. It halts instead, requiring me to keep power button pressed for 4 seconds to actually turn it off after that.
The system in question is Supermicro X9SCAA (http://www.supermicro.com/products/motherboard/ATOM/X9/X9SCAA.cfm). No BIOS (well, UEFI actually) updates are available. With SL6.5 system it DOES power off correctly. So I'm sure there has to be a way to make it work in SL7. But no matter what I tried, I can't seem to make it work. Both with kernel from install and from updates. How I tried to power off: 1) Pressing power button, with default (systemd-based) ACPI power button hook. This calls "systemctl poweroff" and fails eventually. 2) Pressing power button with acpid installed and running (which was required in EL6). The same thing. 3) shutdown -P now 4) poweroff (which is "systemctl poweroff") 5) systemctl poweroff --force --force (doesn't stop services or unmount filesystems, tries to go into S5 state right away but still fails) 6) SysRq + o (supposed to power off through different mechanism and tries to do so but still fails). In any case, the last messages I see on the screen (except for SysRq+o which doesn't print lines about [sda] and doesn't spin down HDD) are Powering off. sd 0:0:0:0: [sda] Synchronizing SCSI cache sd 0:0:0:0: [sda] Stopping disk ACPI: Preparing to enter system sleep state S5 Power down. (It spins down the hard drive and hangs forever. Apparently, it fails to actually enter S5 state) ACPI S5 state is supposed to be supported: # dmesg | grep S5 [ 0.354163] ACPI: (supports S0 S1 S3 S4 S5) I tried lots of various options, including noapic nolapic pci=nocrs (pci=use_crs became default after EL6's 2.6.32 kernel so I thought that maybe it will make the system behave like EL6) pci=noacpi acpi_osi=Linux acpi_osi=! (alone and in all combinations with options above) acpi_osi="Windows 2001" acpi_osi=! (alone and in all combinations with options above) All of above options have their effect, I can see it in dmesg that ACPI and/or PCI devices are initialized differently; but still, none of them help. The bios has options for enabling/disabling S1 & S3 states (these options have effect, but have nothing to do with the problem) and option "Deep S5 sleep state: if Enabled, system supports deep S5 sleep, if Disabled, normal S5 sleep", changing which doesn't affect anything. The somehow related post I found in the internet (the solutions from others like with acpi_osi or pci=noacpi or noapic didn't work for me) was this: https://groups.google.com/forum/#!topic/linux.debian.user/4HkD_7_Ek44 Here the person claims: "It started happening after a kernel upgrade. Before that everything was fine. This is a firmware-related problem. Power down fails, when iternal Intel Graphics Card activated. As a solution - external graphics card." Unfortunately, it's not a solution for me because this is mini-ITX motherboard fit in the small case which supports only low-profile cards and I simply don't have something like low-profile PCI card, nor want to use any add-on cards to keep power consumption low. Even assuming it *is* firmware bug, it works with older kernel. What are the other options to try to make new kernel behave like older one when entering S5 state? Kernel-ml 2.6.13 from elrepo doesn't fix the problem. Full dmesg from default boot (without any options) is here: http://pastebin.com/PqnrVLrd -- Vladimir
