Announcement

Collapse
No announcement yet.

Red Hat Engineer Working On DRM Panic Support For AMDGPU Driver

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • oiaohm
    replied
    Originally posted by Developer12 View Post
    oh I'm sorry, the teeny tiny line saying it was from "debugfs" was too small and lacking in detail to see way down there.

    what the fuck is debug fs, what problem did it have, come on people even windows's new bluescreen is more descriptive than this
    Except there is a problem here your understanding. DRM Panic does not align to Windows Bluescreen of death in modern day versions of windows. DRM panic aligns to current Windows Red Screen of death. Yes the horrible feature of Window Red Screen of death is it common get stuck saying collecting data percentage and never getting to display what in heck has gone wrong at all.


    That message of "test from debugfs" is very detailed in fact.


    This point in the video has the very command that trigger "test from debugfs"
    Code:
    echo 1> /sys/kernel/debug/dri/simple-framebuffer.0/drm_panic_plane_0
    This is not a real kernel panic this is just test the kernel panic code path this is why you can straight up recover from this one without rebooting system.


    Also note here that the line of test is
    "sysrq triggered crash"

    That a "echo c> /proc/sysrq-trigger" or sysrq on keyboard with c yes modern day keyboard alt+print scrreen+c not something you are going to do by mistake.

    Yes every way that you can trigger a crash gives a different line of text. Yes the sysrq is real and can cause file system data loss so don't play with this one unless you have backups.

    Originally posted by Developer12 View Post
    Some text is better than no text. This is nothing but useless abstract art.
    The third line that you did not understand is in fact a detailed answer. The third line is what the Linux kernel panic system thinks is the cause of the issue.

    Yes the QR code on the debugfs message contains enough details to work out what user/process triggered the debugfs bit to go off.

    The windows red screen of death in fact end up less descriptive than this.

    Red screens of death under windows are rare like Linux kernel panics are also quite rare. Recoverable errors are way more common. Yes a windows blue screen of death these days is a recoverable error.

    The QR code contains more text than you can put on the screen otherwise.

    The kmsg option is there have that on screen instead of the short message. Those who have been debugging Linux kernel issues have found without the complete panic message but only having half it it was over 99.99% of the time no better than the single line message. There is a lot of information in that QR code.

    Yes if you get to Windows 11 red screen of death and it displays diagnostic information it only really gives you one line of diagnostic information at most spread out over 3 lines of screen with a instructions will not in fact fix the problem. The reality without the QR code this is no worse than Windows 11 Red Screen of death. With QR code there is more detailed information in the QR code for what went wrong than what the Windows red screen of death provided.

    Leave a comment:


  • Developer12
    replied
    Originally posted by oiaohm View Post

    Also this video
    Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

    Happens to also show the kmsg mode that is the old panic message text dumped to screen. With the classic problem that only part of the message is left displayed on the screen. Yes DRM Panic has a few different configurations.

    No matter what DRM panic is improvement over what use to happen.
    Some text is better than no text. This is nothing but useless abstract art.

    Leave a comment:


  • Developer12
    replied
    Originally posted by oiaohm View Post
    You have three lines of Text and you have only read the first 2.


    Exactly what happen.


    Yes direction to user on the next action the need to do because the computer is not doing anything else other than be solid frozen now due to panic. This can be a count down for kernel configured to auto restart in case of kernel panic. So line is not always displayed its its only display because his kernel did not have the kernel parameter panic=X or echo X>/proc/sys/kernel/panic set yes X is seconds to wait before system auto reboots. So yes this line can be replaced with a count down that count down until the system auto reboots. There is another message that appears if this is a second panic where the system cannot auto reboot because auto reboot has failed.



    This third line is in fact the compact panic message. Yes debugfs is the driver the issue come from and test is the type of issue/error detected from this driver.

    So yes there is a line of test there giving information to start debugging from. Yes this line with a real kernel panic not a triggered one can be a little more detail like segfault at X location(from panic message back-trace) from Y driver.

    Problem is here in the video
    Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

    Yes a Linux kernel full panic report is longer than the number of lines of text that fit on the screen and that only a test report you don't want to see the length of a complex report..
    oh I'm sorry, the teeny tiny line saying it was from "debugfs" was too small and lacking in detail to see way down there.

    what the fuck is debug fs, what problem did it have, come on people even windows's new bluescreen is more descriptive than this

    Leave a comment:


  • oiaohm
    replied
    Originally posted by Developer12 View Post

    OH YES "KERNEL PANIC PLEASE REBOOT YOUR COMPUTER" is SO DESCRIPTIVE.

    And no, a random blob of nonsense pixels is not text. It's a useless abstract art installation.

    This is absolute crap.
    Also this video
    Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

    Happens to also show the kmsg mode that is the old panic message text dumped to screen. With the classic problem that only part of the message is left displayed on the screen. Yes DRM Panic has a few different configurations.

    No matter what DRM panic is improvement over what use to happen.

    Leave a comment:


  • chaotic-entropy
    replied
    I need to stop looking at these comments/forums thinking that I'll learn anything, rather than just see people snapping at each other over nothing.

    Leave a comment:


  • oiaohm
    replied
    Originally posted by Developer12 View Post

    OH YES "KERNEL PANIC PLEASE REBOOT YOUR COMPUTER" is SO DESCRIPTIVE.

    And no, a random blob of nonsense pixels is not text. It's a useless abstract art installation.

    This is absolute crap.
    You have three lines of Text and you have only read the first 2.

    KERNEL PANIC
    Exactly what happen.

    PLEASE REBOOT YOUR COMPUTER
    Yes direction to user on the next action the need to do because the computer is not doing anything else other than be solid frozen now due to panic. This can be a count down for kernel configured to auto restart in case of kernel panic. So line is not always displayed its its only display because his kernel did not have the kernel parameter panic=X or echo X>/proc/sys/kernel/panic set yes X is seconds to wait before system auto reboots. So yes this line can be replaced with a count down that count down until the system auto reboots. There is another message that appears if this is a second panic where the system cannot auto reboot because auto reboot has failed.

    test from debugfs
    This third line is in fact the compact panic message. Yes debugfs is the driver the issue come from and test is the type of issue/error detected from this driver.

    So yes there is a line of test there giving information to start debugging from. Yes this line with a real kernel panic not a triggered one can be a little more detail like segfault at X location(from panic message back-trace) from Y driver.

    Problem is here in the video
    Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

    Yes a Linux kernel full panic report is longer than the number of lines of text that fit on the screen and that only a test report you don't want to see the length of a complex report..

    Leave a comment:


  • Developer12
    replied
    Originally posted by oiaohm View Post

    There is a little bit of text telling you what kind of driver and kind of error.



    QR code contains compressed data. The QR code on the DRM panic screen is the multi page complete kernel panic message compressed. Yes more text than what would fit on the screen the text you would get out a debugging console or the like when the system panics.

    Yes that video compares before and after the DRM Panic screen. Yes its an improvement because before the system just Freezes

    Also note that the QR code compression data allows changing the language of the panic message simply due to the compression method this is another feature useful to developers.

    There are 2 different versions of the DRM Panic screen.
    1) with just text saying what driver caused the panic and type of panic very limited information.
    2) the same text as 1 and the QR code that has the complete kernel panic message including back-trace.

    Yes a common problem of a person using a monitor as debugging console then they photograph the screen/monitor after kernel panic that results in having less than half of the Linux kernel panic message. Lot of ways the kernel panic message has been needing compression.

    The QR code is an improvement for getting the complete kernel panic message. Less than half a kernel panic message is not that useful when you have to basically reproduce the bug to get the same panic again to work out what when wrong and if the fault is highly intermittent or dependent on exact mix of hardware upstream developer may never be able to reproduce this.

    Yes I would like to see a local application for decoding Linux kernel QR code contained data instead of the current go to web site. Yes data security thing.
    OH YES "KERNEL PANIC PLEASE REBOOT YOUR COMPUTER" is SO DESCRIPTIVE.

    And no, a random blob of nonsense pixels is not text. It's a useless abstract art installation.

    This is absolute crap.

    Leave a comment:


  • oiaohm
    replied
    Originally posted by Developer12 View Post

    How about text!? Even the windows bluescreen which they're ripping this off from has plain messages telling you what kind of error is it and what driver it came from.

    Leave it to linux to rip off a user interface and leave out stuff critical to the user experience. It's part of why accessibility is such absolute crap.
    There is a little bit of text telling you what kind of driver and kind of error.



    QR code contains compressed data. The QR code on the DRM panic screen is the multi page complete kernel panic message compressed. Yes more text than what would fit on the screen the text you would get out a debugging console or the like when the system panics.

    Yes that video compares before and after the DRM Panic screen. Yes its an improvement because before the system just Freezes

    Also note that the QR code compression data allows changing the language of the panic message simply due to the compression method this is another feature useful to developers.

    There are 2 different versions of the DRM Panic screen.
    1) with just text saying what driver caused the panic and type of panic very limited information.
    2) the same text as 1 and the QR code that has the complete kernel panic message including back-trace.

    Yes a common problem of a person using a monitor as debugging console then they photograph the screen/monitor after kernel panic that results in having less than half of the Linux kernel panic message. Lot of ways the kernel panic message has been needing compression.

    The QR code is an improvement for getting the complete kernel panic message. Less than half a kernel panic message is not that useful when you have to basically reproduce the bug to get the same panic again to work out what when wrong and if the fault is highly intermittent or dependent on exact mix of hardware upstream developer may never be able to reproduce this.

    Yes I would like to see a local application for decoding Linux kernel QR code contained data instead of the current go to web site. Yes data security thing.

    Leave a comment:


  • access
    replied
    There is text too.

    Leave a comment:


  • Developer12
    replied
    Originally posted by oiaohm View Post

    Not really compare to the historic I had when I was running Nvidia of kernel panic while using X11 with a stack of strange color dots across the top of screen to decode to get the dmesg this is a improvement.

    Lot of ways this is also better than the historic problem of is system just doing a ice station or crashed. Yes something you can still do building too many thing at once and have the interface of Linux completely lag out but will recover this is icestation or has it crashed. Yes the few dots different along the top of screen being the only difference was absolute not great to wake up you have left a system for hour plus because you though it was doing a icestation and still processing but the thing was crashed the complete time.

    DRM Panic support is that you do get a output. QR code on screen where there should not be will be clear the system is dead.

    Some ways if we were being arty have the TUX logo with the QR code in belly for crash message with line under it looks like I ate something bad here.
    How about text!? Even the windows bluescreen which they're ripping this off from has plain messages telling you what kind of error is it and what driver it came from.

    Leave it to linux to rip off a user interface and leave out stuff critical to the user experience. It's part of why accessibility is such absolute crap.

    Leave a comment:

Working...
X