On Sat, Nov 19, 2016 at 8:19 AM, Per Bothner <per@xxxxxxxxxxx> wrote: > > > On 11/18/2016 11:34 PM, Chris Murphy wrote: >> >> On Fri, Nov 18, 2016 at 1:26 PM, Per Bothner <per@xxxxxxxxxxx> wrote: >>> >>> >>> >>> On 11/17/2016 11:05 PM, Chris Murphy wrote: >>>> >>>> >>>> On Thu, Nov 17, 2016 at 7:21 PM, Per Bothner <per@xxxxxxxxxxx> wrote: >>>>> >>>>> >>>>> >>>>> >>>>> On 11/17/2016 02:36 PM, Chris Murphy wrote: >>>>>> >>>>>> >>>>>> >>>>>> Hmm that's been tested quite a bit the last two cycles. I'd sooner >>>>>> expect two USB flash drives have bad cells resulting in corruption on >>>>>> reads than a write corruption bug in mediawriter itself. >>>>> >>>>> >>>>> >>>>> >>>>> I tried a 3rd disk (this one a small USB hard drive), and Live >>>>> booted cleanly! Even better: I got full HD resolution on the >>>>> laptop display and 4K on an external monitor (via my USB docking >>>>> station). >>>> >>>> >>>> >>>> That is an odd duck. Two sticks fail media verification and both give >>>> you fits with display resolution. Can you boot from either of the >>>> sticks, capture dmesg and post it somewhere? >>> >>> >>> >>> This is the screen after "Test this media & start Fedora-Workstation-Live >>> 25_Beta": >>> http://per.bothner.com/tmp/bootfail-F25-B-1-1-20161118.jpg >>> >>> This is the output of dmesg after "Start Fedora-Workstation-Live >>> 25_Beta": >>> http://per.bothner.com/tmp/bootfail-F25-B-1-1-20161118-dmesg.txt >> >> >> That's a 4.8.0rc7 kernel, lots of bug fixes since then and are in the >> final release kernel. > > > I also tried RC 1.3, with the same media test failure. That is what I ended > up actually installing from. > > http://per.bothner.com/tmp/bootfail-F25-1-3.jpg > http://per.bothner.com/tmp/bootfail-F25-1-3-dmesg.txt There are still ACPI errrors there, and I have no idea if they're just noise or a real problem. It might be worth filing bugs upstream for some of these. Thing is, as i'm going through this right now, there are several ACPI related debug options in the kernel, and none of them are enabled on Fedora kernels so to go down this rabbit hole you'd need to be compiling kernels and enable some of this stuff to get more verbose output to attach to the bug reports. I think you said it's a new Spectre, so chances are it doesn't have a firmware update yet. But it's worth checking to make sure. > I just now tried the same flash drive on a older (2-3 years) Toshiba laptop > (with UEFI). > Same media failure. I'm gonna say it's bad media, just because I'm doing the same thing you're doing and can't reproduce it. You could try doing a round of badblocks -w with it, which writes out a bunch of patterns and reads them back in, and will tell you whether it spits back write errors, read errors, or neither but the data is corrupt. Another option is f3 which does a bunch of tests for fake flash, rather exhausitively, and GNOME Multiwriter does an abbreviated fake flash test that should be equally reliable for determining if flash is fake but doesn't fully test the ability of the drive to not corrupt data like f3 and badblocks. Still another idea is to format it with Btrfs and copy a bunch of files to it, and then scrub it. I've definitely experienced flash that corrupts both persistently and transiently. > > So flash drive failures seem most plausible, though it is strange to have > two > identical-mode failures on two different drives. Yes. > > Another possibility is some actual issue that might only be triggered with > older > flash drives, though that seems strange as well. There are bugs in flash firmware, not just the memory chips. There are also bugs in the drivers, especially for new connections. > >> The dmesg shows a bunch of ACPI errors, I can't tell if they're >> critical. Does this laptop have USB-C? > > > Yes: USB-C/Thunderbolt-3. No HDMI port, but it does have a traditional > USB-A port, which I used for booting. Yeah we're in a wonky transitional stage right now. My Spectre, which is not an x360 but rather the thin one, has three USB-C ports. One accepts power, is not Thunderbolt, and is USB 3.1 Gen 1. I can boot Fedora media from it, and get to the installer, and install Fedora without problem. Two USB-C ports are Thunderbolt 3, and are USB 3.1 Gen 2. And while I can boot, strictly being bootloader + kernel + initramfs load, once the kernel looks for devices from which to locate rootfs, it fails. There is no USB stick to be found as far as the kernel is concerned. I still haven't figure out if this is some weird startup bug with the kernel, or if maybe it's an initramfs problem. But from kernel messages, this is really a PCIe port, and it becomes something else at the moment you plug something into it. It's not always necessarily USB. > > At this point, I will just say thanks for the help; I'm good right now; > and the duplicate failures will presumably remain a mystery. Chances are, the behavior will change soon as all of this stuff is in flux, but I still think it's best to gather as much data about each individual problem and get it upstream just because they probably don't have this particular hardware handy. -- Chris Murphy _______________________________________________ test mailing list -- test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to test-leave@xxxxxxxxxxxxxxxxxxxxxxx
