[IA64] Support domU coredump on ia64

This patch supports domU coredump on ia64.  xen_panic_event() is
registered to panic_notifier_list, and xen_panic_event() calls
HYPERVISOR_shutdown(SHUTDOWN_crash) at panic time.

If xend is notified of crash status, xend calls dumpCore()
and create domU's core in /var/xen/dump.

For sample crash module and usage, see:

http://lists.xensource.com/archives/html/xen-ia64-devel/2006-07/msg00230.html

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>
[minor code re-arrangement]
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Fix fetch code method when FP fault occurs @VTi side

This patch intends to use __vmx_get_domain_bundle to fetch code
when FP fault @VTi side.

Singed-off-by: Zhang xiantao <xiantao.zhang@intel.com>

[IA64] vDSO paravirtualization: paravirtualize vDSO

paravirtualize vdso areabased on Kevin's pointout and Dan's Idea.
introduce hyperprivop HYPERPRIVOP_RSM_BE and HYPERPRIVOP_GET_PSR.
and paravirtualize vdso area using them.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] vDSO paravirtualization: import linux files

import gate.S, gate.ld.S and patch.c which are needed to
paravirtualize vdso area.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] ivt.S cleanup

Cleanup: unused code #ifndef XEN-ized, some insns swapped to improve
bundling.

Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>

[IA64] Revert xen-ia64-unstable.hg cset 10685

This reverts the following xen-ia64-unstable.hg cset:

"[IA64] emulate PAL_HALT_LIGHT on domU"
10685:8ad37880564dff8599f2f410f7bd0c54a50a29e2

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] optimize entry and exit path

VMM save/restore r4~r7 and unat to handle virtualization fault
and mmio access, but it is not necessary for other faults to
save/restore these registers.  This patch is to save/restore
these registers only when dtlb and virtualization fault happen.

Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[IA64] xencons command line parameter is now an integer

'xencons' command line is now an integer, so that it can be disabled.

Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>

[IA64] optimize vpsr

vpsr can't keep track flowing bits of guest psr
be,up,ac,mfl,mfh,cpl,ri.
Previously every time xen gets control, xen will sync
vpsr with cr.ipsr, it's not neccessary.
Xen sync with cr.ipsr when needed.

Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[IA64] do not purge vhpt when emulation itr

It's safe for linux not to purge vhpt when emulating itr,
otherwise there is considerable performance loss

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Zhang xiantao <xiantao.zhang@intel.com>
Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[IA64] enable acceleration of external interrupt

This patch is to enable acceleration of externel interrupt
which is described in VTI spec.

Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[IA64] change default to use xencons, disable dom0 serial driver

This enables xencons by default.  Serial console users should no longer
need to specify creative values for xenconsole= and console= on the
boot command line.  By default, xencons will use ttyS0, so make sure
dom0 has a getty running on that device.

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] fix usage of ia64_xen_vector

Must look up the vector via the GSI to test if it's in use.

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Save & restore.

xc_ia64_linux_save.c and xc_ia64_linux_restore.c added.
vcpu context has more registers and states (eg: tr registers).
Per cpu irqs are deallocated when cpu is switched off.
#if/#endif added in reboot.c for ia64.

Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>

[IA64] Creates tools/libxc/ia64 directory.

Split and move xc_ia64_stubs.c into ia64/xc_ia64_hvm_build.c and
ia64/xc_ia64_stubs.c
Creates ia64/Makefile.
Clean up (static in const) in xc_ia64_hvm_build.c

Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>
[whitespace cleanups in new files]
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] don't change NaTPAGE into WB

Xen/IA64 changes memory attributes of tlb into WB to prevent domU from
uncachable access.  But NaTPage attribute need not to be changed to WB.
Linux uses NaTPage attributes.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
`

[IA64] remove duplicated console_endboot().

There are two. one is in construct_dom0() and another is in start_kernel()
Like x86, left the one in start_kernel()

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] fix vcpu_itr_i(), vcpu_itr_d() and vcpu_flush_tlb_vhpt_range() callers

- vcpu_itr_i() and vcpu_itr_d() must purge vTLB entry which overlaps the
  new inserted entry.
- some address argument of vcpu_flush_tlb_vhpt_range() was wrong.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] clean up linux-xen/tlb.c compiler warnings

Signed-off-by: Al Stone <ahs3@fc.hp.com>

[IA64] clean up pal_emul.c compiler warnings

This patch cleans up some compiler warnings in pal_emul.c.  The
problem was that a struct ia64_pal_retval was being used to return
the results from emulated PAL calls, but not all of the fields
were being initialized.  Given that fields were not initialized,
I was also thinking that there might be a remote chance of data
leaking between vcpu's -- but I haven't convinced myself that
that could actually happen.

Signed-off-by: Al Stone <ahs3@fc.hp.com>
[whitespace and line wrapping cleanup throughout file]
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] emulate PAL_HALT_LIGHT on domU

This patch emulates Guest PAL_HALT_LIGHT on domU by using do_block and timer.
It also adds the function of the timer interrupt to domU at the vcpu woke up.

Signed-off-by: Atsushi SAKAI <sakaia@jp.fujitsu.com>
[warning fixes and static inlining]
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] replace lost linux-null files

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] merge in INIT patches

[IA64] Enable FW accleration for VTi

Signed-off-by: Zhang xiantao <xiantao.zhang@intel.com>

[IA64] separate mapped_regs and vpd, handle various page size

Allow Xen page size != 16KB.

Decorelate XSI size and page_size.
Separate mapped_regs and vpd to save memory in non-VTi mode.

Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>

[IA64] Force make symbolic link when make xen

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] INIT hadler for support coredumping feature

This patch is for supporting coredumping feature.
Some dump feature is necessary to save registers into memory.
So I implemented it and add a hook of calling dump function.
Because all cpu have to save registers,
the same handler is called by monarch and slave handler.

This patch is useful not only dump feature, but also debug xen.
(Because INIT hadler is always able to show register and call trace.)

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>
[Updated for linux-xen location]
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Add mca_asm, mca into Makefile

This patch turn on new init handler.

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>
[Updated for new mca.c location]
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Modify mca.c for Xen/ia64 INIT support

Update mca.c to support INIT on Xen

In ia64_mca_cpu_init(), use alloc_xenheap_pages() for allocation
of ia64_mca_cpu.  In linux, alloc_bootmem is used.

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>
[Ported from stand-alone mca.c]
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Add mca.c

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] #ifdef out unused mca_asm.S code

Make ia64_os_mca_dispatch spin

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>
[Changed from deleting code to #ifdef'ing it out]
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Add mca_asm.S to linux-xen

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Fix MINSTATE_GET_CURRENT

To get virtual curent address in both physcal mode and virtual mode,
add MINSTATE_GET_CURRENT_VIRT.

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] Fix MINSTATE_START/END_SAVE_MIN_PHYS for INIT handler

- THIS_CPU(ia64_mca_data)  have physcal address of each cpu's ia64_mca_cpu.
  I computed address of init_stack by using it.
- from dep r12=-1,r12,61,3; to dep r12=-1,r12,60,4;
  to computed xen virtual address.

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] add unw_init_from_interruption()

From stock upstream unwind.c

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>
[fixed patch to match upstream]
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Fix LOAD_PHYSCAL macro

This fix is workaround. Now LOAD_PHYSCAL is used only
by MCA/INIT handler.

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] Move asmmacro.h to linux-xen

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Fix GET_THIS_PADDR

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] Fix GATE_ADDR for ia64_do_show_stack()

fix GATE_ADDR and define it in include/asm-ia64/xensystem.h.

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] add IA64_MCA_CPU_INIT_STACK_OFFSET

add IA64_MCA_CPU_INIT_STACK_OFFSET which is used in
MINSTATE_START_SAVE_MIN_PHYS

Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] ia64 softirq clean ups.

This is the counter part of the C/S
10565:53f552ad404234c457fdd62560c9e8b0ea976674
Xen softirq is only checked only before returning to domain context.
So that it's non-sense to check in_interrupt() and add/sub SOFTIRQ_OFFSET.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] build warning fix

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] Bug fix: DOM0_GETMEMLIST: do not return -EINVAL in case of success!

Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>

merge with xen-unstable.hg

[XEN] Add CPUID hypervisor-info leaves at index 0x40000000.
Currently only a signature leaf is defined ("Xen\0").
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] getpagframeinfo field name mfn -> gmfn.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] rewrite rfi emulation

This approach of emulating rfi is straightforward

Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[XEN] Hypercall-init dom0_op takes GMFN, not MFN.
Signed-off-by: Steven Smith <sos22@cl.cam.ac.uk>

[IA64] fix domain_flush_vtlb_all()

It should purge software tlb entry of specified vcpu, not current.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] more tlb flush when vcpu migration between pcpu.

When vcpu is migrated to another pcpu and then it returns to the first
pcpu, VHPT and mTLB must also be flushed.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] Implement and use DOM0_DOMAIN_STEUP.

DOM0_GETMEMLIST now reads ptes and use gpfn.
Domain builder reworked: calls DOMAIN_SETUP, setup start_info page.
SAL data are now in domain memory.
is_vti field added in domain.arch.
Many cleanups (indentation, static, warnings).

Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>

[HVM][VMX] More descriptive failed vmentry.
Signed-off-by: Xin Li <xin.b.li@intel.com>

[MINIOS] Mapping page frames on demand added to the memory management.

Signed-off-by: Steven Smith <sos22@cam.ac.uk>
Signed-off-by: Grzegorz Milos <gm281@cam.ac.uk>

[MINIOS]Mapping page frames on demand added to the memory management.
Signed-off-by: Steven Smith <sos22@cam.ac.uk>
Signed-off-by: Grzegorz Milos <gm281@cam.ac.uk>

[XEN] Separate domain creation from vcpu creation.
Creating a domain no longer creates vcpu0 -- that is now
done later.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XENMON] This patch removes the magic number "31" for readability.
The number "31" means the idle domain ID.

In detail:
- display the idle domain ID with "Idle" instead of "31"
- write to the file "log-idle.log" instead of "log-dom31.log".

Signed-off-by: KUWAMURA Shin'ya <kuwa@jp.fujitsu.com>

[HVM] Fix the kernel build failure issue on HVM SMP guests which
are using 2M PSE pages. Now we can do kernel build on:
 1) RHEL4 IA32 SMP guest on PAE host.
 2) RHEL4 IA32 SMP guest on IA32e host.
 3) IA32e SMP guest on IA32e host.

Signed-off-by: Xiaohui Xin <xiaohui.xin@intel.com>
Signed-off-by: Xin Li <xin.b.li@intel.com>

[MINIOS]Events handling cleaned up. The interface extended to provide
void* pointer to handlers.
Signed-off-by: Steven Smith <sos22@cam.ac.uk>
Signed-off-by: Grzegorz Milos <gm281@cam.ac.uk>

[MINIOS] XenBus tests disabled in MiniOS. Xend bug induced
by write to device/vif/0.
Signed-off-by: Grzegorz Milos <gm281@cam.ac.uk>

[MINIOS] Implement XenBus transactions in MiniOS.
Signed-off-by: Steven Smith <sos22@cam.ac.uk>

[XENTOP] Fix Field attributes and width.

It miss that set the FIELD attribute and width for VBD
information in xentop.c.

Signed-off-by  Satoshi UCHIDA <s-uchida@ap.jp.nec.com>

[HVM][VMX] Move vmcs and I/O bitmap allocation into
vmx_initialise_guest_resources().

Signed-off-by: Xin B Li <xin.b.li@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Update xen caps to indicate that HVM IA32 PAE guest on PAE host is available.
Signed-off-by: Xin Li <xin.b.li@intel.com>

[HVM][VMX] expose PSE feature to HVM guests if PAE is turned on in
config file, with this change, RHEL4 IA32 SMP guest can boot.
But because we don't support 4Mbytes PSE pages, so x86_32 HVM guest
can NOT use PSE feature. However user may have "pae=1" in config file,
which causes PSE feature seen by x86_32 HVM guest, and it will then use
4Mbytes PSE pages. In this case, we will have to crash x86_32 HVM.
So when running x86_32 HVM guest, user need pay attention not to have
"pae=1" in config file.

Signed-off-by: Xin Li <xin.b.li@intel.com>
Signed-off-by: Xiaohui Xin <xiaohui.xin@intel.com>

[IA64] add the definition of __cacheline_aligned and its variants

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[HVM] Disable pmtimer in qemu acpi device model. Needs fixing
to not burn ~25% of a 3GHz CPU.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] fix console_start_sync() undefined warning

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] allow vcpu to move between pcpus

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] sync console before reset

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] A few clean ups in __xen_create/destroy_contiguous_region.

This is the IA64 counter part of the C/S
10563:a75c1bdfe76146ad591574d805be5be0a3098d2a of xen-unstable.hg

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] Clean up unused functions

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[IA64] Emulate PAL_HALT_LIGHT

Use do_block to emulate Guest PAL_HALT_LIGHT,
whenever there is an interrupt for this vcpu,
this vcpu is woken up.

Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[IA64] update VTI Readme file

Update VTI how-to readme file with physical disk

Signed-off-by: Fred yang <Fred.Yang@intel.com>

[IA64] privop_stat.c

Cleanup: create privop_stat.[ch]
Move all stats function to privop_stat.c to depolute other files.

Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>

[IA64] Fix a typo.

Signed-off-by: Zhang xiantao <xiantao.zhang@intel.com>

[NET]: Rename feature-tso to feature-gso-tcpv4

This patch renames the name feature-tso to feature-gso-tcpv4 for future
expansion.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

[NET] front: Zero negotiated bits in xen_set_features

When we reconnect to the backend we need to first zero all negotiated
bits as the functions xen_set_sg and xen_set_tso do not (and are not
supposed to) zero bits when they fail to set them.

This patch also permanently enables the NETIF_F_GSO_ROBUST bit as we
never parse any GSO fields ourselves (even if we did the backend could
not trust us so it's wasted effort).

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

[NET] back: Add GSO features field and check gso_size

This patch adds the as-yet unused GSO features which will contain
protocol-independent bits such as the ECN marker.

It also makes the backend check gso_size to ensure that it is non-zero.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

[NET] back: Fix off-by-one error in netbk_tx_err

The generalised extra request info patch introduced a bug with the use
of netbk_tx_err since it advanced the req_cons pointer by one.  This
patch fixes thing by delaying the increment in netbk_tx_err.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

[BUILD] Miscellaneous build-system fixes and tweaks.

* Don't delete non-existent %.tar.bz2 in buildconfigs/Rules.mk's %-mrproper
* Add a mrpropper target to buildconfigs/mk.linux-2.6-xen, and
  have this delete the local tarball
  - This should also delete the local patch, if there is one, but I am not
    sure how to do this with the current incarntation of the code, and I
    will post a subsequent patch which makes this easier and effects this
    chang. As there currently is no local patch this is not a big deal at
    this moment. (Horms)
* Stop kclean from erroring out if the kernel directory doesn't exist
* Remove patches/*/.makedep in top level Makefile rather than in
  buildconfigs/Rules.mk, as calling rules in the latter causes
  patches/*/.make to be created (Magnus)

Signed-Off-By: Magnus Damm <magnus@valinux.co.jp>
Signed-Off-By: Horms <horms@verge.net.au>

[DOC] Add descriptions on mouse handling and how to use the USB
support that is now in the HVM device model.

Signed-off-by: Don Dugger <donald.d.dugger@intel.com>

[LINUX] Fix booting with e.g. xencons=tty3, so that tty1/tty2 are valid (dummy) devices.

Attached patch makes it always create valid devices on tty1-tty63, but
all but the one specified by xencons are dummy devices.

From: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Merge with xen-ia64-unstable.hg

[HVM][VMX] Fixes to VMCS create/destroy.
Signed-off-by: Keir Fraser <keir@xensource.com>

[LIBXC] Fix the 64-bit build.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Sync Elf parsing with libxc parser.
Signed-off-by: Keir Fraser <keir@xensource.com>

Provide more specific and helpful error messages during ELF loading.

Signed-off-by: Jimi Xenidis <jimix@watson.ibm.com>
Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

Allow 32-bit libxc to load 64-bit ELF files.
- use 64-bit integral types for addresses in struct domain_start_info
- use stroull() to parse 64-bit values
- remove redundant _p(a) definition and add a comment
- printf format changes for the new types

Signed-off-by: Jimi Xenidis <jimix@watson.ibm.com>
Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

[XEN] Mark two more opcodes (C6 and C7) as Mov instructions.
From: Mats Petersson <Mats.Petersson@amd.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[XENBUS] Another simplification -- bus_for_each_dev() better than bus_find_device().
Signed-off-by: Keir Fraser <keir@xensource.com>

[XENBUS] Another fix to the wait-for-device-connection code.
The new method for printing unconnected device status is arguably
cleaner and also does not screw up reference counts (previous code
called bus_find_device() with no put_device() on the result.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XENBUS] Do not wait for devices with no driver to connect.
Signed-off-by: Keir Fraser <keir@xensource.com>

Merge

Don't modify global active weight when changing weight of inactive domain.
Signed-off-by: Emmanuel Ackaouy <ack@xensource.com>

[XENBUS] Improve the code for waiting for devices to connect. Provide
a more useful error when devices fail to connect.
From: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[NET]: Update net-gso.patch. Remove net-tso.patch.

New changeset merged upstream:

    [TCP]: Reset gso_segs if packet is dodgy

    I wasn't paranoid enough in verifying GSO information.  A bogus gso_segs
    could upset drivers as much as a bogus header would.  Let's reset it in
    the per-protocol gso_segment functions.

    I didn't verify gso_size because that can be verified by the source of
    the dodgy packets.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

[BLK] back: Only remove sysfs attributes if they were previously
added. Also check for errors during attribute creation.
Signed-off-by: Keir Fraser <keir@xensource.com>

Remove use of unlikely -- it's not defined on all systems.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

Revert 10521, 10526 and 10527. This completes reversion of ACM modifications to xm.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Various softirq cleanups. Main one is to always
call smp_processor_id() after any softirq, as rescheduling
may cause us to move to another processor on ia64
(spotted by Isaku Yamahata). Also get rid of many direct
callers of do_softirq() by creating new function
process_pending_timers().
Signed-off-by: Keir Fraser <keir@xensource.com>