On 28/12/2018 12:04, Roger Pau Monne wrote: > Hello, > > This series includes some miscellaneous fixes for the pv-shim mode, > specially regarding the handling of the memory map.
Unfortunately, something overall is now wonky. With this series applied, I'm getting: (d9) (XEN) Initial PVH-e820 RAM map: (d9) (XEN) 0000000000000000 - 00000000fc000000 (usable) (d9) (XEN) 00000000fc000000 - 00000000fc009040 (ACPI data) (d9) (XEN) 00000000feff8000 - 00000000ff000000 (reserved) (d9) (XEN) 0000000100000000 - 0000000104000400 (usable) (d9) (XEN) Checking MTRR ranges... (d9) (XEN) MTRR cap: 508 type: 806 (d9) (XEN) PVH-e820 RAM map: (d9) (XEN) 0000000000000000 - 00000000000a0000 (usable) (d9) (XEN) 00000000fc000000 - 00000000fc009040 (ACPI data) (d9) (XEN) 00000000feff8000 - 00000000ff000000 (reserved) (d9) (XEN) 0000000100000000 - 0000000104000400 (usable) which a) is trying to account for the legacy range despite starting as a PVH guest. (I think that perhaps cpu_has_hypervisor doesn't work that early, and I'll try making a fix for it.) b) The entire remainder of the RAM block which covered the legacy hole got deleted, so Xen failed to relocate itself and panic()'d c) Xen really oughtn't to relocate itself in cases like this, but that is way out of scope here. ~Andrew _______________________________________________ Xen-devel mailing list Xen-devel@lists.xenproject.org https://lists.xenproject.org/mailman/listinfo/xen-devel
