Mailing List Archive

[PATCH v2 0/5] x86: Remove more 16M total-size restrictions
The boot/directmap pagetables, high Xen pagetables, and use of BOOT_FS all
come with arbitrary limitations to Xen's total size. Remove these limits.

Testing of the EFI build indicates that there is still an issue lurking

(XEN) ----[ Xen-4.14-unstable x86_64 debug=y Not tainted ]----
(XEN) CPU: 0
(XEN) RIP: e008:[<ffff82d0802557fa>] drivers/char/ns16550.c#ns16550_poll+0x1d/0x21
(XEN) Xen call trace:
(XEN) [<ffff82d0802557fa>] R drivers/char/ns16550.c#ns16550_poll+0x1d/0x21
(XEN) [<ffff82d080247446>] F common/timer.c#execute_timer+0x49/0x64
(XEN) [<ffff82d080247c24>] F common/timer.c#timer_softirq_action+0xa2/0x276
(XEN) [<ffff82d080243b81>] F common/softirq.c#__do_softirq+0x71/0x9a
(XEN) [<ffff82d080243bdd>] F process_pending_softirqs+0x33/0x37
(XEN) [<ffff82d0812bd1be>] F __cpu_up+0x652/0x719
(XEN) [<ffff82d080204874>] F cpu_up+0x75/0xe3
(XEN) [<ffff82d08162ad43>] F __start_xen+0x251a/0x29b1
(XEN) ****************************************
(XEN) Panic on CPU 0:
(XEN) FATAL TRAP: vector = 6 (invalid opcode)
(XEN) ****************************************

which is run_in_exception_handler() not being recognised as BUGFRAME_run_fn.
Therefore, I've left the linker assert in place for now.

Andrew Cooper (5):
x86/boot: Create the l2_xenmap[] mappings dynamically
x86/boot: Size the boot/directmap mappings dynamically
x86/boot: Drop explicit %fs uses
x86/boot: Simplify pagetable manipulation loops
x86/boot: Drop sym_fs()

xen/arch/x86/boot/head.S | 145 +++++++++++++++++++++++------------------
xen/arch/x86/boot/trampoline.S | 1 -
xen/arch/x86/boot/x86_64.S | 23 +++----
xen/arch/x86/efi/efi-boot.h | 37 +++++++++--
xen/arch/x86/ | 3 +
5 files changed, 126 insertions(+), 83 deletions(-)


Xen-devel mailing list