Linux Kernel Live Dump
| Název práce v češtině: | Dump z běžícího linuxového jádra |
|---|---|
| Název v anglickém jazyce: | Linux Kernel Live Dump |
| Klíčová slova: | crashdump|Linux|debugování|postcopy |
| Klíčová slova anglicky: | crashdump|Linux|debugging|postcopy |
| Akademický rok vypsání: | 2022/2023 |
| Typ práce: | diplomová práce |
| Jazyk práce: | angličtina |
| Ústav: | Katedra distribuovaných a spolehlivých systémů (32-KDSS) |
| Vedoucí / školitel: | Mgr. Michal Koutný |
| Řešitel: | skrytý - zadáno a potvrzeno stud. odd. |
| Datum přihlášení: | 07.03.2023 |
| Datum zadání: | 08.03.2023 |
| Datum potvrzení stud. oddělením: | 24.10.2023 |
| Oponenti: | Mgr. Martin Děcký, Ph.D. |
| Konzultanti: | prof. Ing. Petr Tůma, Dr. |
| Zásady pro vypracování |
| The Linux kernel has the ability to dump state of the system when a fatal error is encountered and the system can no longer execute. The dump provides exhaustive information that helps determine the likely cause of the crash.
The goal of the thesis is to extend the dump functionality to "live dumps", that is, the ability to dump state at arbitrary moments in system execution. Such dumps can be useful when troubleshooting non fatal issues, where crashes do not happen or where reboots are too intrusive. Live dumps are already possible for guest kernels under a hypervisor coordination, however, we are primarily focusing on bare metal setups and setups with confidential virtual machines, where live dumps are not available. An ideal dump should contain a CPU and memory state snapshot that corresponds to what a simultaneous interrupt during classic kernel panic observes, without stopping all CPUs for the duration of the dump (which may not be feasible on production systems with real memory sizes and target device throughputs). The thesis should not deviate from this goal in a way that would interfere with the debugging purpose of the dump. Additionally, the dumping mechanism could also serve as a vehicle for live migration (focus on confidential VMs). Possible approaches are similar to the dumps during live VM migration (under a hypervisor) or modified suspend to disk. These ideas should not be limiting, broader analysis is also expected. The implementation (but not necessarily the analysis) should target the mainline Linux kernel on x86_64. The implementation should be accompanied with an evaluation of the dump impact and restrictions on common workloads. |
| Seznam odborné literatury |
| [1] https://www.kernel.org/doc/html/latest/admin-guide/kdump/kdump.html
[2] https://lwn.net/Kernel/Index/#Crash_dumps [3] M. Hines, U. Deshpande, and K. Gopalan. Post-copy live migration of virtual machines. In SIGOPS Operating Systems Review, July 2009. [4] https://research.ibm.com/publications/secure-live-migration-of-encrypted-vms [5] Linux Kernel Mailing List archive, https://lore.kernel.org/lkml/ |