sdk/emulator/qemu.git
9 years agotypofixes - v4
Veres Lajos [Tue, 8 Sep 2015 21:45:14 +0000 (22:45 +0100)]
typofixes - v4

Signed-off-by: Veres Lajos <vlajos@gmail.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agoqapi-schema: remove legacy<> from doc
Marc-André Lureau [Fri, 4 Sep 2015 19:41:01 +0000 (21:41 +0200)]
qapi-schema: remove legacy<> from doc

The legacy<> type is no longer used since 7ce7ffe02.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agodisas/microblaze: Remove unused code
Stefan Weil [Sat, 29 Aug 2015 07:44:33 +0000 (09:44 +0200)]
disas/microblaze: Remove unused code

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agohelp: dd missing newline
Laurent Vivier [Fri, 4 Sep 2015 19:30:04 +0000 (21:30 +0200)]
help: dd missing newline

Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agoTarget-ppc: Remove unnecessary variable
Shraddha Barke [Fri, 4 Sep 2015 19:20:28 +0000 (00:50 +0530)]
Target-ppc: Remove unnecessary variable

Compress lines and remove the variable ret.

Change made using Coccinelle script

@@
expression ret;
@@
- if (ret) return ret;
- return 0;
+ return ret;
@@
local idexpression ret;
expression e;
@@
- ret = e;
- return ret;
+ return e;
@@
type T; identifier i;
@@
- T i;
... when != i

Signed-off-by: Shraddha Barke <shraddha.6596@gmail.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agobaum: Fix build with debugging enabled
Samuel Thibault [Sun, 30 Aug 2015 15:12:13 +0000 (17:12 +0200)]
baum: Fix build with debugging enabled

cur and buf are pointers, so the difference is a ptrdiff_t

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agolinux-user: Fix warnings caused by missing 'static' attribute
Stefan Weil [Sat, 29 Aug 2015 07:29:52 +0000 (09:29 +0200)]
linux-user: Fix warnings caused by missing 'static' attribute

Warnings from the Sparse static analysis tool:

linux-user/main.c:40:12: warning:
 symbol 'filename' was not declared. Should it be static?
linux-user/main.c:41:12: warning:
 symbol 'argv0' was not declared. Should it be static?
linux-user/main.c:42:5: warning:
 symbol 'gdbstub_port' was not declared. Should it be static?
linux-user/main.c:43:11: warning:
 symbol 'envlist' was not declared. Should it be static?

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agoopts: produce valid command line in qemu_opts_print
Kővágó, Zoltán [Tue, 7 Jul 2015 14:42:10 +0000 (16:42 +0200)]
opts: produce valid command line in qemu_opts_print

This will let us print options in a format that the user would actually
write it on the command line (foo=bar,baz=asd,etc=def), without
prepending a spurious comma at the beginning of the list, or quoting
values unnecessarily.  This patch provides the following changes:
* write and id=, if the option has an id
* do not print separator before the first element
* do not quote string arguments
* properly escape commas (,) for QEMU

Signed-off-by: Kővágó, Zoltán <DirtY.iCE.hu@gmail.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agodocs: fix a qga/qapi-schema.json comment
Marc-André Lureau [Thu, 27 Aug 2015 10:48:35 +0000 (12:48 +0200)]
docs: fix a qga/qapi-schema.json comment

For consistency with the rest of the comment blocks.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agotrivial: remove trailing newline from error_report
John Snow [Mon, 29 Jun 2015 20:56:26 +0000 (16:56 -0400)]
trivial: remove trailing newline from error_report

Minor cleanup.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agomaint: avoid useless "if (foo) free(foo)" pattern
Markus Armbruster [Wed, 26 Aug 2015 12:02:53 +0000 (14:02 +0200)]
maint: avoid useless "if (foo) free(foo)" pattern

My Coccinelle semantic patch finds a few more, because it also fixes up
the equally pointless conditional

    if (foo) {
        free(foo);
        foo = NULL;
    }

Result (feel free to squash it into your patch):

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agomaint: avoid useless "if (foo) free(foo)" pattern
Daniel P. Berrange [Wed, 26 Aug 2015 11:17:18 +0000 (12:17 +0100)]
maint: avoid useless "if (foo) free(foo)" pattern

The free() and g_free() functions both happily accept
NULL on any platform QEMU builds on. As such putting a
conditional 'if (foo)' check before calls to 'free(foo)'
merely serves to bloat the lines of code.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agomaint: remove unused include for strings.h
Daniel P. Berrange [Wed, 26 Aug 2015 11:17:17 +0000 (12:17 +0100)]
maint: remove unused include for strings.h

A number of files were including strings.h but not using any
of the functions it provides

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agomaint: remove unused include for signal.h
Daniel P. Berrange [Wed, 26 Aug 2015 11:17:16 +0000 (12:17 +0100)]
maint: remove unused include for signal.h

A number of files were including signal.h but not using any
of the functions it provides

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agomaint: remove unused include for dirent.h
Daniel P. Berrange [Wed, 26 Aug 2015 11:17:15 +0000 (12:17 +0100)]
maint: remove unused include for dirent.h

A number of files were including dirent.h but not using any
of the functions it provides

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agomaint: remove unused include for assert.h
Daniel P. Berrange [Wed, 26 Aug 2015 11:17:14 +0000 (12:17 +0100)]
maint: remove unused include for assert.h

A number of files were including assert.h but not using any
of the functions it provides

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agomaint: remove / fix many doubled words
Daniel P. Berrange [Wed, 26 Aug 2015 11:17:13 +0000 (12:17 +0100)]
maint: remove / fix many doubled words

Many source files have doubled words (eg "the the", "to to",
and so on). Most of these can simply be removed, but a couple
were actual mis-spellings (eg "to to" instead of "to do").
There was even one triple word score "to to to" :-)

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agomaint: remove double semicolons in many files
Daniel P. Berrange [Wed, 26 Aug 2015 11:17:12 +0000 (12:17 +0100)]
maint: remove double semicolons in many files

A number of source files have statements accidentally
terminated by a double semicolon - eg 'foo = bar;;'.
This is harmless but a mistake none the less.

The tcg/ia64/tcg-target.c file is whitelisted because
it has valid use of ';;' in a comment containing assembly
code.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agoi6300esb: fix timer overflow
Laurent Vivier [Tue, 4 Aug 2015 08:27:31 +0000 (10:27 +0200)]
i6300esb: fix timer overflow

We use muldiv64() to compute the time to wait:

    timeout = muldiv64(get_ticks_per_sec(), timeout, 33000000);

but get_ticks_per_sec() is 10^9 (30 bit value) and timeout
is a 35 bit value.

Whereas muldiv64 is:

    uint64_t muldiv64(uint64_t a, uint32_t b, uint32_t c)

So we loose 3 bits of timeout.

Swapping get_ticks_per_sec() and timeout fixes it.

We can also replace it by a multiplication by 30 ns,
but this changes PCI clock frequency from 33MHz to 33.333333MHz
and we need to do this on all the QEMU PCI devices (later...)

Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agoTrivial: fix commandline help message
Andrey Korolyov [Fri, 31 Jul 2015 19:12:54 +0000 (22:12 +0300)]
Trivial: fix commandline help message

Fix obvious typo in printed help for qemu-nbd.

Signed-off-by: Andrey Korolyov <andrey@xdel.ru>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agoUpdate language files for QEMU 2.4.0
Stefan Weil [Thu, 30 Jul 2015 05:46:43 +0000 (07:46 +0200)]
Update language files for QEMU 2.4.0

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
9 years agoMerge remote-tracking branch 'remotes/sstabellini/tags/xen-2015-09-10-tag' into staging
Peter Maydell [Thu, 10 Sep 2015 17:25:52 +0000 (18:25 +0100)]
Merge remote-tracking branch 'remotes/sstabellini/tags/xen-2015-09-10-tag' into staging

xen-2015-09-10

# gpg: Signature made Thu 10 Sep 2015 17:52:08 BST using RSA key ID 70E1AE90
# gpg: Good signature from "Stefano Stabellini <stefano.stabellini@eu.citrix.com>"

* remotes/sstabellini/tags/xen-2015-09-10-tag: (29 commits)
  xen/pt: Don't slurp wholesale the PCI configuration registers
  xen/pt: Check for return values for xen_host_pci_[get|set] in init
  xen/pt: Move bulk of xen_pt_unregister_device in its own routine.
  xen/pt: Make xen_pt_unregister_device idempotent
  xen/pt: Log xen_host_pci_get/set errors in MSI code.
  xen/pt: Log xen_host_pci_get in two init functions
  xen/pt: Remove XenPTReg->data field.
  xen/pt: Check if reg->init function sets the 'data' past the reg->size
  xen/pt: Sync up the dev.config and data values.
  xen/pt: Use xen_host_pci_get_[byte|word] instead of dev.config
  xen/pt: Use XEN_PT_LOG properly to guard against compiler warnings.
  xen/pt/msi: Add the register value when printing logging and error messages
  xen: use errno instead of rc for xc_domain_add_to_physmap
  xen/pt: xen_host_pci_config_read returns -errno, not -1 on failure
  xen/pt: Make xen_pt_msi_set_enable static
  xen/pt: Update comments with proper function name.
  xen/HVM: atomically access pointers in bufioreq handling
  xen-hvm: When using xc_domain_add_to_physmap also include errno when reporting
  xen, gfx passthrough: add opregion mapping
  xen, gfx passthrough: register host bridge specific to passthrough
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoxen/pt: Don't slurp wholesale the PCI configuration registers
Konrad Rzeszutek Wilk [Wed, 8 Jul 2015 19:58:41 +0000 (15:58 -0400)]
xen/pt: Don't slurp wholesale the PCI configuration registers

Instead we have the emulation registers ->init functions which
consult the host values to see what the initial value should be
and they are responsible for populating the dev.config.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Check for return values for xen_host_pci_[get|set] in init
Konrad Rzeszutek Wilk [Thu, 2 Jul 2015 18:33:44 +0000 (14:33 -0400)]
xen/pt: Check for return values for xen_host_pci_[get|set] in init

and if we have failures we call xen_pt_destroy introduced in
'xen/pt: Move bulk of xen_pt_unregister_device in its own routine.'
and free all of the allocated structures.

Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Move bulk of xen_pt_unregister_device in its own routine.
Konrad Rzeszutek Wilk [Tue, 8 Sep 2015 20:21:59 +0000 (16:21 -0400)]
xen/pt: Move bulk of xen_pt_unregister_device in its own routine.

This way we can call it if we fail during init.

This code movement introduces no changes.

Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Make xen_pt_unregister_device idempotent
Konrad Rzeszutek Wilk [Tue, 8 Sep 2015 20:21:29 +0000 (16:21 -0400)]
xen/pt: Make xen_pt_unregister_device idempotent

To deal with xen_host_pci_[set|get]_ functions returning error values
and clearing ourselves in the init function we should make the
.exit (xen_pt_unregister_device) function be idempotent in case
the generic code starts calling .exit (or for fun does it before
calling .init!).

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Log xen_host_pci_get/set errors in MSI code.
Konrad Rzeszutek Wilk [Wed, 24 Jun 2015 21:27:40 +0000 (17:27 -0400)]
xen/pt: Log xen_host_pci_get/set errors in MSI code.

We seem to only use these functions when de-activating the
MSI - so just log errors.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Log xen_host_pci_get in two init functions
Konrad Rzeszutek Wilk [Wed, 24 Jun 2015 21:18:26 +0000 (17:18 -0400)]
xen/pt: Log xen_host_pci_get in two init functions

To help with troubleshooting in the field.

Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Remove XenPTReg->data field.
Konrad Rzeszutek Wilk [Wed, 1 Jul 2015 19:41:33 +0000 (15:41 -0400)]
xen/pt: Remove XenPTReg->data field.

We do not want to have two entries to cache the guest configuration
registers: XenPTReg->data and dev.config. Instead we want to use
only the dev.config.

To do without much complications we rip out the ->data field
and replace it with an pointer to the dev.config. This way we
have the type-checking (uint8_t, uint16_t, etc) and as well
and pre-computed location.

Alternatively we could compute the offset in dev.config by
using the XenPTRRegInfo and XenPTRegGroup every time but
this way we have the pre-computed values.

This change also exposes some mis-use:
 - In 'xen_pt_status_reg_init' we used u32 for the Capabilities Pointer
   register, but said register is an an u16.
 - In 'xen_pt_msgdata_reg_write' we used u32 but should have only use u16.

Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Check if reg->init function sets the 'data' past the reg->size
Konrad Rzeszutek Wilk [Mon, 29 Jun 2015 20:41:14 +0000 (16:41 -0400)]
xen/pt: Check if reg->init function sets the 'data' past the reg->size

It should never happen, but in case it does (an developer adds
a new register and the 'init_val' expands past the register
size) we want to report. The code will only write up to
reg->size so there is no runtime danger of the register spilling
across other ones - however to catch this sort of thing
we still return an error.

Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Sync up the dev.config and data values.
Konrad Rzeszutek Wilk [Mon, 29 Jun 2015 20:24:40 +0000 (16:24 -0400)]
xen/pt: Sync up the dev.config and data values.

For a passthrough device we maintain a state of emulated
registers value contained within d->config. We also consult
the host registers (and apply ro and write masks) whenever
the guest access the registers. This is done in xen_pt_pci_write_config
and xen_pt_pci_read_config.

Also in this picture we call pci_default_write_config which
updates the d->config and if the d->config[PCI_COMMAND] register
has PCI_COMMAND_MEMORY (or PCI_COMMAND_IO) acts on those changes.

On startup the d->config[PCI_COMMAND] are the host values, not
what the guest initial values should be, which is exactly what
we do _not_ want to do for 64-bit BARs when the guest just wants
to read the size of the BAR. Huh you say?

To get the size of 64-bit memory space BARs,  the guest has
to calculate ((BAR[x] & 0xFFFFFFF0) + ((BAR[x+1] & 0xFFFFFFFF) << 32))
which means it has to do two writes of ~0 to BARx and BARx+1.

prior to this patch and with XSA120-addendum patch (Linux kernel)
the PCI_COMMAND register is copied from the host it can have
PCI_COMMAND_MEMORY bit set which means that QEMU will try to
update the hypervisor's P2M with BARx+1 value to ~0 (0xffffffff)
(to sync the guest state to host) instead of just having
xen_pt_pci_write_config and xen_pt_bar_reg_write apply the
proper masks and return the size to the guest.

To thwart this, this patch syncs up the host values with the
guest values taking into account the emu_mask (bit set means
we emulate, PCI_COMMAND_MEMORY and PCI_COMMAND_IO are set).
That is we copy the host values - masking out any bits which
we will emulate. Then merge it with the initial emulation register
values. Lastly this value is then copied both in
dev.config _and_ XenPTReg->data field.

There is also reg->size accounting taken into consideration
that ends up being used in patch.
 xen/pt: Check if reg->init function sets the 'data' past the reg->size

This fixes errors such as these:

(XEN) memory_map:add: dom2 gfn=fffe0 mfn=fbce0 nr=20
(DEBUG) 189 pci dev 04:0 BAR16 wrote ~0.
(DEBUG) 200 pci dev 04:0 BAR16 read 0x0fffe0004.
(XEN) memory_map:remove: dom2 gfn=fffe0 mfn=fbce0 nr=20
(DEBUG) 204 pci dev 04:0 BAR16 wrote 0x0fffe0004.
(DEBUG) 217 pci dev 04:0 BAR16 read upper 0x000000000.
(XEN) memory_map:add: dom2 gfn=ffffffff00000 mfn=fbce0 nr=20
(XEN) p2m.c:883:d0v0 p2m_set_entry failed! mfn=ffffffffffffffff rc:-22
(XEN) memory_map:fail: dom2 gfn=ffffffff00000 mfn=fbce0 nr=20 ret:-22
(XEN) memory_map:remove: dom2 gfn=ffffffff00000 mfn=fbce0 nr=20
(XEN) p2m.c:920:d0v0 gfn_to_mfn failed! gfn=ffffffff00000 type:4
(XEN) p2m.c:920:d0v0 gfn_to_mfn failed! gfn=ffffffff00001 type:4
..
(XEN) memory_map: error -22 removing dom2 access to [fbce0,fbcff]
(DEBUG) 222 pci dev 04:0 BAR16 read upper 0x0ffffffff.
(XEN) memory_map:remove: dom2 gfn=ffffffff00000 mfn=fbce0 nr=20
(XEN) memory_map: error -22 removing dom2 access to [fbce0,fbcff]

[The DEBUG is to illustate what the hvmloader was doing]

Also we swap from xen_host_pci_long to using xen_host_pci_get_[byte,word,long].

Otherwise we get:

xen_pt_config_reg_init: Offset 0x0004 mismatch! Emulated=0x0000, host=0x2300017, syncing to 0x2300014.
xen_pt_config_reg_init: Error: Offset 0x0004:0x2300014 expands past register size(2)!

which is not surprising. We read the value as an 32-bit (from host),
then operate it as a 16-bit - and the remainder is left unchanged.

We end up writing the value as 16-bit (so 0014) to dev.config
(as we use proper xen_set_host_[byte,word,long] so we don't spill
to other registers) but in XenPTReg->data it is as 32-bit (0x2300014)!

It is harmless as the read/write functions end up using an size mask
and never modify the bits past 16-bit (reg->size is 2).

This patch fixes the warnings by reading the value using the
proper size.

Note that the check for size is still left in-case the developer
sets bits past the reg->size in the ->init routines. The author
tried to fiddle with QEMU_BUILD_BUG to make this work but failed.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Reported-by: Sander Eikelenboom <linux@eikelenboom.it>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Use xen_host_pci_get_[byte|word] instead of dev.config
Konrad Rzeszutek Wilk [Mon, 29 Jun 2015 18:01:13 +0000 (14:01 -0400)]
xen/pt: Use xen_host_pci_get_[byte|word] instead of dev.config

During init time we treat the dev.config area as a cache
of the host view. However during execution time we treat it
as guest view (by the generic PCI API). We need to sync Xen's
code to the generic PCI API view. This is the first step
by replacing all of the code that uses dev.config or
pci_get_[byte|word] to get host value to actually use the
xen_host_pci_get_[byte|word] functions.

Interestingly in 'xen_pt_ptr_reg_init' we also needed to swap
reg_field from uint32_t to uint8_t - since the access is only
for one byte not four bytes. We can split this as a seperate
patch however we would have to use a cast to thwart compiler
warnings in the meantime.

We also truncated 'flags' to 'flag' to make the code fit within
the 80 characters.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoMerge remote-tracking branch 'remotes/armbru/tags/pull-error-2015-09-10' into staging
Peter Maydell [Thu, 10 Sep 2015 13:51:35 +0000 (14:51 +0100)]
Merge remote-tracking branch 'remotes/armbru/tags/pull-error-2015-09-10' into staging

error: On abort, report where the error was created

# gpg: Signature made Thu 10 Sep 2015 13:01:39 BST using RSA key ID EB918653
# gpg: Good signature from "Markus Armbruster <armbru@redhat.com>"
# gpg:                 aka "Markus Armbruster <armbru@pond.sub.org>"

* remotes/armbru/tags/pull-error-2015-09-10:
  error: On abort, report where the error was created
  error: Revamp interface documentation
  error: error_set_errno() is unused, drop
  qga/vss-win32: Document the DLL requires non-null errp
  qga: Clean up unnecessarily dirty casts
  error: Make error_setg() a function
  error: De-duplicate code creating Error objects

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoxen/pt: Use XEN_PT_LOG properly to guard against compiler warnings.
Konrad Rzeszutek Wilk [Mon, 29 Jun 2015 20:06:19 +0000 (16:06 -0400)]
xen/pt: Use XEN_PT_LOG properly to guard against compiler warnings.

If XEN_PT_LOGGING_ENABLED is enabled the XEN_PT_LOG macros start
using the first argument. Which means if within the function there
is only one user of the argument ('d') and XEN_PT_LOGGING_ENABLED
is not set, we get compiler warnings. This is not the case now
but with the "xen/pt: Use xen_host_pci_get_[byte|word] instead of dev.config"
we will hit - so this sync up the function to the rest of them.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt/msi: Add the register value when printing logging and error messages
Konrad Rzeszutek Wilk [Mon, 29 Jun 2015 16:30:37 +0000 (12:30 -0400)]
xen/pt/msi: Add the register value when printing logging and error messages

We would like to know what the MSI register value is to help
in troubleshooting in the field. As such modify the logging
logic to include such details in xen_pt_msgctrl_reg_write.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen: use errno instead of rc for xc_domain_add_to_physmap
Konrad Rzeszutek Wilk [Mon, 29 Jun 2015 16:51:05 +0000 (12:51 -0400)]
xen: use errno instead of rc for xc_domain_add_to_physmap

In Xen 4.6 commit cd2f100f0f61b3f333d52d1737dd73f02daee592
"libxc: Fix do_memory_op to return negative value on errors"
made the libxc API less odd-ball: On errors, return value is
-1 and error code is in errno. On success the return value
is either 0 or an positive value.

Since we could be running with an old toolstack in which the
Exx value is in rc or the newer, we add an wrapper around
the xc_domain_add_to_physmap (called xen_xc_domain_add_to_physmap)
which will always return the EXX.

Xen 4.6 did not change the libxc functions mentioned (same parameters)
so we piggyback on the fact that Xen 4.6 has a new function:
commit 504ed2053362381ac01b98db9313454488b7db40 "tools/libxc: Expose
new hypercall xc_reserved_device_memory_map" and check for that.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Suggested-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: xen_host_pci_config_read returns -errno, not -1 on failure
Konrad Rzeszutek Wilk [Mon, 29 Jun 2015 17:58:17 +0000 (13:58 -0400)]
xen/pt: xen_host_pci_config_read returns -errno, not -1 on failure

However the init routines assume that on errors the return
code is -1 (as the libxc API is) - while those xen_host_* routines follow
another paradigm - negative errno on return, 0 on success.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Make xen_pt_msi_set_enable static
Konrad Rzeszutek Wilk [Wed, 24 Jun 2015 21:26:43 +0000 (17:26 -0400)]
xen/pt: Make xen_pt_msi_set_enable static

As we do not use it outside our code.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/pt: Update comments with proper function name.
Konrad Rzeszutek Wilk [Wed, 24 Jun 2015 21:16:01 +0000 (17:16 -0400)]
xen/pt: Update comments with proper function name.

It has changed but the comments still refer to the old names.

Reviewed-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen/HVM: atomically access pointers in bufioreq handling
Jan Beulich [Fri, 24 Jul 2015 09:38:28 +0000 (03:38 -0600)]
xen/HVM: atomically access pointers in bufioreq handling

The number of slots per page being 511 (i.e. not a power of two) means
that the (32-bit) read and write indexes going beyond 2^32 will likely
disturb operation. The hypervisor side gets I/O req server creation
extended so we can indicate that we're using suitable atomic accesses
where needed, allowing it to atomically canonicalize both pointers when
both have gone through at least one cycle.

The Xen side counterpart (which is not a functional prereq to this
change, albeit a build one) went in already (commit b7007bc6f9).

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen-hvm: When using xc_domain_add_to_physmap also include errno when reporting
Konrad Rzeszutek Wilk [Fri, 13 Mar 2015 19:36:58 +0000 (15:36 -0400)]
xen-hvm: When using xc_domain_add_to_physmap also include errno when reporting

.errors - as it will most likely have the proper error value.

Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen, gfx passthrough: add opregion mapping
Tiejun Chen [Wed, 15 Jul 2015 05:37:50 +0000 (13:37 +0800)]
xen, gfx passthrough: add opregion mapping

The OpRegion shouldn't be mapped 1:1 because the address in the host
can't be used in the guest directly.

This patch traps read and write access to the opregion of the Intel
GPU config space (offset 0xfc).

The original patch is from Jean Guyader <jean.guyader@eu.citrix.com>

Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen, gfx passthrough: register host bridge specific to passthrough
Tiejun Chen [Wed, 15 Jul 2015 05:37:49 +0000 (13:37 +0800)]
xen, gfx passthrough: register host bridge specific to passthrough

Just register that pci host bridge specific to passthrough.

Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen, gfx passthrough: register a isa bridge
Tiejun Chen [Wed, 15 Jul 2015 05:37:48 +0000 (13:37 +0800)]
xen, gfx passthrough: register a isa bridge

Currently we just register this isa bridge when we use IGD
passthrough in Xen side.

Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoigd gfx passthrough: create a isa bridge
Tiejun Chen [Wed, 15 Jul 2015 05:37:47 +0000 (13:37 +0800)]
igd gfx passthrough: create a isa bridge

Currently IGD drivers always need to access PCH by 1f.0. But we
don't want to poke that directly to get ID, and although in real
world different GPU should have different PCH. But actually the
different PCH DIDs likely map to different PCH SKUs. We do the
same thing for the GPU. For PCH, the different SKUs are going to
be all the same silicon design and implementation, just different
features turn on and off with fuses. The SW interfaces should be
consistent across all SKUs in a given family (eg LPT). But just
same features may not be supported.

Most of these different PCH features probably don't matter to the
Gfx driver, but obviously any difference in display port connections
will so it should be fine with any PCH in case of passthrough.

So currently use one PCH version, 0x8c4e, to cover all HSW(Haswell)
scenarios, 0x9cc3 for BDW(Broadwell).

Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
9 years agoxen, gfx passthrough: retrieve VGA BIOS to work
Tiejun Chen [Wed, 15 Jul 2015 05:37:46 +0000 (13:37 +0800)]
xen, gfx passthrough: retrieve VGA BIOS to work

Now we retrieve VGA bios like kvm stuff in qemu but we need to
fix Device Identification in case if its not matched with the
real IGD device since Seabios is always trying to compare this
ID to work out VGA BIOS.

Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen, gfx passthrough: basic graphics passthrough support
Tiejun Chen [Wed, 15 Jul 2015 05:37:45 +0000 (13:37 +0800)]
xen, gfx passthrough: basic graphics passthrough support

basic gfx passthrough support:
- add a vga type for gfx passthrough
- register/unregister legacy VGA I/O ports and MMIOs for passthrough GFX

Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Signed-off-by: Yang Zhang <yang.z.zhang@Intel.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agohw/pci-assign: split pci-assign.c
Tiejun Chen [Wed, 15 Jul 2015 05:37:44 +0000 (13:37 +0800)]
hw/pci-assign: split pci-assign.c

We will try to reuse assign_dev_load_option_rom in xen side, and
especially its a good beginning to unify pci assign codes both on
kvm and xen in the future.

[Fix build for Windows]

Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
9 years agopiix: create host bridge to passthrough
Tiejun Chen [Wed, 15 Jul 2015 05:37:43 +0000 (13:37 +0800)]
piix: create host bridge to passthrough

Implement a pci host bridge specific to passthrough. Actually
this just inherits the standard one. And we also just expose
a minimal real host bridge pci configuration subset.

[Replace pread with lseek and read to fix Windows build]

Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
9 years agoerror: On abort, report where the error was created
Markus Armbruster [Fri, 19 Jun 2015 17:21:59 +0000 (19:21 +0200)]
error: On abort, report where the error was created

This is particularly useful when we abort in error_propagate(),
because there the stack backtrace doesn't lead to where the error was
created.  Looks like this:

    Unexpected error in parse_block_error_action() at .../qemu/blockdev.c:322:
    qemu-system-x86_64: -drive if=none,werror=foo: 'foo' invalid write error action
    Aborted (core dumped)

Note: to get this example output, I monkey-patched drive_new() to pass
&error_abort to blockdev_init().

To keep the error handling boiler plate from growing even more, all
error_setFOO() become macros expanding into error_setFOO_internal()
with additional __FILE__, __LINE__, __func__ arguments.  Not exactly
pretty, but it works.

The macro trickery breaks down when you take the address of an
error_setFOO().  Fortunately, we do that in just one place: qemu-ga's
Windows VSS provider and requester DLL wants to call
error_setg_win32() through a function pointer "to avoid linking glib
to the DLL".  Use error_setg_win32_internal() there.  The use of the
function pointer is already wrapped in a macro, so the churn isn't
bad.

Code size increases by some 35KiB for me (0.7%).  Tolerable.  Could be
less if we passed relative rather than absolute source file names to
the compiler, or forwent reporting __func__.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Acked-by: Laszlo Ersek <lersek@redhat.com>
9 years agoerror: Revamp interface documentation
Markus Armbruster [Fri, 19 Jun 2015 16:29:24 +0000 (18:29 +0200)]
error: Revamp interface documentation

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agoerror: error_set_errno() is unused, drop
Markus Armbruster [Fri, 19 Jun 2015 20:16:14 +0000 (22:16 +0200)]
error: error_set_errno() is unused, drop

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agoqga/vss-win32: Document the DLL requires non-null errp
Markus Armbruster [Sat, 20 Jun 2015 07:33:56 +0000 (09:33 +0200)]
qga/vss-win32: Document the DLL requires non-null errp

requester.cpp uses this pattern to receive an error and pass it on to
the caller (err_is_set() macro peeled off for clarity):

    ... code that may set errset->errp ...
    if (errset->errp && *errset->errp) {
        ... handle error ...
    }

This breaks when errset->errp is null.  As far as I can tell, it
currently isn't, so this is merely fragile, not actually broken.

The robust way to do this is to receive the error in a local variable,
then propagate it up, like this:

    Error *err = NULL;

    ... code that may set err ...
    if (err)
        ... handle error ...
        error_propagate(errset->errp, err);
    }

See also commit 5e54769, 0f230bf, a903f40.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agoqga: Clean up unnecessarily dirty casts
Markus Armbruster [Fri, 19 Jun 2015 18:44:54 +0000 (20:44 +0200)]
qga: Clean up unnecessarily dirty casts

qga_vss_fsfreeze() casts error_set_win32() from

    void (*)(Error **, int, ErrorClass, const char *, ...)

to

    void (*)(void **, int, int, const char *, ...)

The result is later called.  Since the two types are not compatible,
the call is undefined behavior.  It works in practice anyway.

However, there's no real need for trickery here.  Clean it up as
follows:

* Declare struct Error, and fix the first parameter.

* Switch to error_setg_win32().  This gets rid of the troublesome
  ErrorClass parameter.  Requires converting error_setg_win32() from
  macro to function, but that's trivially easy, because this is the
  only user of error_set_win32().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agoerror: Make error_setg() a function
Markus Armbruster [Fri, 19 Jun 2015 13:36:16 +0000 (15:36 +0200)]
error: Make error_setg() a function

Saves a tiny amount of code at every call site.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agoerror: De-duplicate code creating Error objects
Markus Armbruster [Fri, 19 Jun 2015 11:59:47 +0000 (13:59 +0200)]
error: De-duplicate code creating Error objects

Duplicated when commit 680d16d added error_set_errno(), and again when
commit 20840d4 added error_set_win32().

Make the original copy in error_set() reusable by factoring out
error_setv(), then rewrite error_set_errno() and error_set_win32() on
top of it.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging
Peter Maydell [Thu, 10 Sep 2015 09:24:30 +0000 (10:24 +0100)]
Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging

virtio,pc,acpi fixes, cleanups

Fixes all over the place.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# gpg: Signature made Thu 10 Sep 2015 10:16:18 BST using RSA key ID D28D5469
# gpg: Good signature from "Michael S. Tsirkin <mst@kernel.org>"
# gpg:                 aka "Michael S. Tsirkin <mst@redhat.com>"

* remotes/mst/tags/for_upstream:
  hw/pci: fix pci_update_mappings() trace events
  pc: memhotplug: keep reserved-memory-end broken on 2.4 and earlier machines
  pc: memhotplug: fix incorrectly set reserved-memory-end
  acpi: Remove unused definition.
  virtio: avoid leading underscores for helpers
  pc: Remove redundant arguments from xen_hvm_init()
  pci: Fix pci_device_iommu_address_space() bus propagation

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agohw/pci: fix pci_update_mappings() trace events
Laszlo Ersek [Tue, 1 Sep 2015 21:33:23 +0000 (23:33 +0200)]
hw/pci: fix pci_update_mappings() trace events

The current trace prototypes and (matching) trace calls lead to
"unorthodox" PCI BDF notation in at least the stderr trace backend. For
example, the four BARs of a QXL video card at 00:01.0 (bus 0, slot 1,
function 0) are traced like this (PID and timestamps removed):

  pci_update_mappings_add d=0x7f14a73bf890 00:00.1 0,0x84000000+0x4000000
  pci_update_mappings_add d=0x7f14a73bf890 00:00.1 1,0x80000000+0x4000000
  pci_update_mappings_add d=0x7f14a73bf890 00:00.1 2,0x88200000+0x2000
  pci_update_mappings_add d=0x7f14a73bf890 00:00.1 3,0xd060+0x20

The slot and function values are in reverse order.

Stick with the conventional BDF notation.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Don Koch <dkoch@verizon.com>
Cc: qemu-trivial@nongnu.org
Fixes: 7828d75045
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agopc: memhotplug: keep reserved-memory-end broken on 2.4 and earlier machines
Igor Mammedov [Mon, 7 Sep 2015 11:55:32 +0000 (13:55 +0200)]
pc: memhotplug: keep reserved-memory-end broken on 2.4 and earlier machines

it will prevent guests on old machines from seeing
inconsistent memory mapping in firmware/ACPI views.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
9 years agopc: memhotplug: fix incorrectly set reserved-memory-end
Igor Mammedov [Mon, 7 Sep 2015 11:55:31 +0000 (13:55 +0200)]
pc: memhotplug: fix incorrectly set reserved-memory-end

reserved-memory-end tells firmware address from which
it could start treating memory as PCI address space
and map PCI BARs after it to avoid collisions with
RAM.
Currently it is incorrectly pointing to address where
hotplugged memory range starts which could redirect
hotplugged RAM accesses to PCI BARs when firmware
maps them over RAM or viceverse.
Fix this by pointing reserved-memory-end to the end
of memory hotplug area.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
9 years agoacpi: Remove unused definition.
Richard W.M. Jones [Wed, 2 Sep 2015 19:03:38 +0000 (20:03 +0100)]
acpi: Remove unused definition.

Signed-off-by: Richard W.M. Jones <rjones@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agovirtio: avoid leading underscores for helpers
Cornelia Huck [Mon, 17 Aug 2015 09:48:29 +0000 (11:48 +0200)]
virtio: avoid leading underscores for helpers

Commit ef546f1275f6563e8934dd5e338d29d9f9909ca6 ("virtio: add
feature checking helpers") introduced a helper __virtio_has_feature.
We don't want to use reserved identifiers, though, so let's
rename __virtio_has_feature to virtio_has_feature and virtio_has_feature
to virtio_vdev_has_feature.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agopc: Remove redundant arguments from xen_hvm_init()
Eduardo Habkost [Mon, 17 Aug 2015 18:42:29 +0000 (11:42 -0700)]
pc: Remove redundant arguments from xen_hvm_init()

Remove arguments that can be found in PCMachineState.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agopci: Fix pci_device_iommu_address_space() bus propagation
Benjamin Herrenschmidt [Sat, 4 Jul 2015 23:19:15 +0000 (09:19 +1000)]
pci: Fix pci_device_iommu_address_space() bus propagation

he current code walks up the bus tree for an iommu, however it passes
to the iommu_fn() callback the bus/devfn of the immediate child of
the level where the callback was found, rather than the original
bus/devfn where the search started from.

This prevents iommu's like POWER8 (and in fact also Q35) to properly
provide an address space for a subset of devices that aren't immediate
children of the iommu.

PCIe carries the originator bdfn acccross to the iommu on all DMA
transactions, so we must be able to properly identify devices at all
levels.

This changes the function pci_device_iommu_address_space() to pass
the original pointers to the iommu_fn() callback instead.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20150908' into...
Peter Maydell [Tue, 8 Sep 2015 17:02:36 +0000 (18:02 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20150908' into staging

target-arm queue:
 * Implement priority handling properly via GICC_APR
 * Enable TZ extensions on the GIC if we're using them
 * Minor preparatory patches for EL3 support
 * cadence_gem: Correct Marvell PHY SPCFC reset value
 * Support AHCI in ZynqMP

# gpg: Signature made Tue 08 Sep 2015 17:48:33 BST using RSA key ID 14360CDE
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>"
# gpg:                 aka "Peter Maydell <pmaydell@gmail.com>"
# gpg:                 aka "Peter Maydell <pmaydell@chiark.greenend.org.uk>"

* remotes/pmaydell/tags/pull-target-arm-20150908:
  xlnx-zynqmp: Connect the sysbus AHCI to ZynqMP
  xlnx-zynqmp.c: Convert some of the error_propagate() calls to error_abort
  ahci.c: Don't assume AHCIState's parent is AHCIPCIState
  ahci: Separate the AHCI state structure into the header
  cadence_gem: Correct Marvell PHY SPCFC reset value
  target-arm: Add AArch64 access to PAR_EL1
  target-arm: Correct opc1 for AT_S12Exx
  target-arm: Log the target EL when taking exceptions
  target-arm: Fix default_exception_el() function for the case when EL3 is not supported
  hw/arm/virt: Enable TZ extensions on the GIC if we are using them
  hw/arm/virt: Default to not providing TrustZone support
  hw/cpu/{a15mpcore, a9mpcore}: enable TrustZone in GIC if it is enabled in CPUs
  hw/intc/arm_gic_common: Configure IRQs as NS if doing direct NS kernel boot
  hw/arm: new interface for devices which need to behave differently for kernel boot
  qom: Add recursive version of object_child_for_each
  hw/intc/arm_gic: Actually set the active bits for active interrupts
  hw/intc/arm_gic: Drop running_irq and last_active arrays
  hw/intc/arm_gic: Fix handling of GICC_APR<n>, GICC_NSAPR<n> registers
  hw/intc/arm_gic: Running priority is group priority, not full priority
  armv7m_nvic: Implement ICSR without using internal GIC state

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoxlnx-zynqmp: Connect the sysbus AHCI to ZynqMP
Alistair Francis [Tue, 8 Sep 2015 16:38:45 +0000 (17:38 +0100)]
xlnx-zynqmp: Connect the sysbus AHCI to ZynqMP

Connect the Sysbus AHCI device to ZynqMP.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Sai Pavan Boddu <saipava@xilinx.com>
[PMM: removed unnecessary brackets in error_propagate call]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoxlnx-zynqmp.c: Convert some of the error_propagate() calls to error_abort
Alistair Francis [Tue, 8 Sep 2015 16:38:45 +0000 (17:38 +0100)]
xlnx-zynqmp.c: Convert some of the error_propagate() calls to error_abort

Convert all of the non-realize error_propagate() calls into error_abort
calls as they shouldn't be user visible failure cases.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoahci.c: Don't assume AHCIState's parent is AHCIPCIState
Alistair Francis [Tue, 8 Sep 2015 16:38:45 +0000 (17:38 +0100)]
ahci.c: Don't assume AHCIState's parent is AHCIPCIState

The AHCIState struct can either have AHCIPCIState or SysbusAHCIState
as a parent. The ahci_irq_lower() and ahci_irq_raise() functions
assume that it is always AHCIPCIState, which is not always the
case, which causes a seg fault. Verify what the container of AHCIState
is before setting the PCIDevice struct.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Acked-by: John Snow <jsnow@redhat.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoahci: Separate the AHCI state structure into the header
Alistair Francis [Tue, 8 Sep 2015 16:38:45 +0000 (17:38 +0100)]
ahci: Separate the AHCI state structure into the header

Pull the AHCI state structure out into the header. This allows
other containers to access the struct. This is required to add
the device to modern SoC containers.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Sai Pavan Boddu <saipava@xilinx.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agocadence_gem: Correct Marvell PHY SPCFC reset value
Alistair Francis [Tue, 8 Sep 2015 16:38:45 +0000 (17:38 +0100)]
cadence_gem: Correct Marvell PHY SPCFC reset value

Bit 15 of the PHY Specific Status Register is reserved and
should remain 0. Fix the reset value to ensure that the 15th
bit is not set.

Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: c795069e49040ff770fe2ece19dfe1791b729e22.1441316450.git.alistair.francis@xilinx.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add AArch64 access to PAR_EL1
Edgar E. Iglesias [Tue, 8 Sep 2015 16:38:44 +0000 (17:38 +0100)]
target-arm: Add AArch64 access to PAR_EL1

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Message-id: 1441311266-8644-4-git-send-email-edgar.iglesias@gmail.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Correct opc1 for AT_S12Exx
Edgar E. Iglesias [Tue, 8 Sep 2015 16:38:44 +0000 (17:38 +0100)]
target-arm: Correct opc1 for AT_S12Exx

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Message-id: 1441311266-8644-3-git-send-email-edgar.iglesias@gmail.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Log the target EL when taking exceptions
Edgar E. Iglesias [Tue, 8 Sep 2015 16:38:44 +0000 (17:38 +0100)]
target-arm: Log the target EL when taking exceptions

Log the target EL when taking exceptions. This is useful when
debugging guest SW or QEMU itself while transitioning through
the various ELs.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Message-id: 1441311266-8644-2-git-send-email-edgar.iglesias@gmail.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Fix default_exception_el() function for the case when EL3 is not supported
Sergey Sorokin [Tue, 8 Sep 2015 16:38:44 +0000 (17:38 +0100)]
target-arm: Fix default_exception_el() function for the case when EL3 is not supported

If EL3 is not supported in current configuration,
we should not try to get EL3 bitness.

Signed-off-by: Sergey Sorokin <afarallax@yandex.ru>
Message-id: 1441208342-10601-2-git-send-email-afarallax@yandex.ru
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agohw/arm/virt: Enable TZ extensions on the GIC if we are using them
Peter Maydell [Tue, 8 Sep 2015 16:38:44 +0000 (17:38 +0100)]
hw/arm/virt: Enable TZ extensions on the GIC if we are using them

If we're creating a board with support for TrustZone, then enable
it on the GIC model as well as on the CPUs.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1441383782-24378-7-git-send-email-peter.maydell@linaro.org

9 years agohw/arm/virt: Default to not providing TrustZone support
Peter Maydell [Tue, 8 Sep 2015 16:38:44 +0000 (17:38 +0100)]
hw/arm/virt: Default to not providing TrustZone support

Switch the default for the 'virt' board to not providing TrustZone
support in either the CPU or the GIC. This is primarily for the
benefit of UEFI, which currently assumes there is no TrustZone
support, and does not set the GIC up correctly if it is TZ-aware.
It also means the board is consistent about its behaviour whether
we're using KVM or TCG (KVM never has TrustZone support).

If TrustZone support is required (for instance for running test
suites or TZ-aware firmware) it can be enabled with the
"-machine secure=on" command line option.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1441383782-24378-6-git-send-email-peter.maydell@linaro.org

9 years agohw/cpu/{a15mpcore, a9mpcore}: enable TrustZone in GIC if it is enabled in CPUs
Peter Maydell [Tue, 8 Sep 2015 16:38:43 +0000 (17:38 +0100)]
hw/cpu/{a15mpcore, a9mpcore}: enable TrustZone in GIC if it is enabled in CPUs

If the A9 and A15 CPUs which we're creating the peripherals for have
TrustZone (EL3) enabled, then also enable it in the GIC we create.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1441383782-24378-5-git-send-email-peter.maydell@linaro.org

9 years agohw/intc/arm_gic_common: Configure IRQs as NS if doing direct NS kernel boot
Peter Maydell [Tue, 8 Sep 2015 16:38:43 +0000 (17:38 +0100)]
hw/intc/arm_gic_common: Configure IRQs as NS if doing direct NS kernel boot

If we directly boot a kernel in NonSecure on a system where the GIC
supports the security extensions then we must cause the GIC to
configure its interrupts into group 1 (NonSecure) rather than the
usual group 0, and with their initial priority set to the highest
NonSecure priority rather than the usual highest Secure priority.
Otherwise the guest kernel will be unable to use any interrupts.

Implement this behaviour, controlled by a flag which we set if
appropriate when the ARM bootloader code calls our ARMLinuxBootIf
interface callback.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1441383782-24378-4-git-send-email-peter.maydell@linaro.org

9 years agohw/arm: new interface for devices which need to behave differently for kernel boot
Peter Maydell [Tue, 8 Sep 2015 16:38:43 +0000 (17:38 +0100)]
hw/arm: new interface for devices which need to behave differently for kernel boot

For ARM we have a little minimalist bootloader in hw/arm/boot.c which
takes the place of firmware if we're directly booting a Linux kernel.
Unfortunately a few devices need special case handling in this situation
to do the initialization which on real hardware would be done by
firmware. (In particular if we're booting a kernel in NonSecure state
then we need to make a TZ-aware GIC put all its interrupts into Group 1,
or the guest will be unable to use them.)

Create a new QOM interface which can be implemented by devices which
need to do something different from their default reset behaviour.
The callback will be called after machine initialization and before
first reset.

Suggested-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1441383782-24378-3-git-send-email-peter.maydell@linaro.org

9 years agoqom: Add recursive version of object_child_for_each
Peter Crosthwaite [Tue, 8 Sep 2015 16:38:43 +0000 (17:38 +0100)]
qom: Add recursive version of object_child_for_each

Useful for iterating through an entire QOM subtree.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1441383782-24378-2-git-send-email-peter.maydell@linaro.org

9 years agohw/intc/arm_gic: Actually set the active bits for active interrupts
Peter Maydell [Tue, 8 Sep 2015 16:38:43 +0000 (17:38 +0100)]
hw/intc/arm_gic: Actually set the active bits for active interrupts

Although we were correctly handling interrupts becoming active
and then inactive, we weren't actually exposing this to the guest
by setting the 'active' flag for the interrupt, so reads
of GICD_ICACTIVERn and GICD_ISACTIVERn would generally incorrectly
return zeroes. Correct this oversight.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1438089748-5528-6-git-send-email-peter.maydell@linaro.org

9 years agohw/intc/arm_gic: Drop running_irq and last_active arrays
Peter Maydell [Tue, 8 Sep 2015 16:38:42 +0000 (17:38 +0100)]
hw/intc/arm_gic: Drop running_irq and last_active arrays

The running_irq and last_active arrays represent state which
doesn't exist in a real hardware GIC. The only thing we use
them for is updating the running priority when an interrupt
is completed, but in fact we can use the active-priority
registers to do this. The running priority is always the
priority corresponding to the lowest set bit in the active
priority registers, because only one interrupt at any
particular priority can be active at once.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1438089748-5528-5-git-send-email-peter.maydell@linaro.org

9 years agohw/intc/arm_gic: Fix handling of GICC_APR<n>, GICC_NSAPR<n> registers
Peter Maydell [Tue, 8 Sep 2015 16:38:42 +0000 (17:38 +0100)]
hw/intc/arm_gic: Fix handling of GICC_APR<n>, GICC_NSAPR<n> registers

A GICv2 has both GICC_APR<n> and GICC_NSAPR<n> registers, with
the latter holding the active priority bits for Group 1 interrupts
(usually Nonsecure interrupts), and the Nonsecure view of the
GICC_APR<n> is the second half of the GICC_NSAPR<n> registers.
Turn our half-hearted implementation of APR<n> into a proper
implementation of both APR<n> and NSAPR<n>:

 * Add the underlying state for NSAPR<n>
 * Make sure APR<n> aren't visible for pre-GICv2
 * Implement reading of NSAPR<n>
 * Make non-secure reads of APR<n> behave correctly
 * Implement writing to APR<n> and NSAPR<n>

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1438089748-5528-4-git-send-email-peter.maydell@linaro.org

9 years agohw/intc/arm_gic: Running priority is group priority, not full priority
Peter Maydell [Tue, 8 Sep 2015 16:38:42 +0000 (17:38 +0100)]
hw/intc/arm_gic: Running priority is group priority, not full priority

Priority values for the GIC are divided into a "group priority"
and a "subpriority" (with the division being determined by the
binary point register). The running priority is only determined
by the group priority of the active interrupts, not the
subpriority. In particular, this means that there can't be more
than one active interrupt at any particular group priority.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1438089748-5528-3-git-send-email-peter.maydell@linaro.org

9 years agoarmv7m_nvic: Implement ICSR without using internal GIC state
Peter Maydell [Tue, 8 Sep 2015 16:38:42 +0000 (17:38 +0100)]
armv7m_nvic: Implement ICSR without using internal GIC state

Change the implementation of the Interrupt Control and State Register
in the v7M NVIC to not use the running_irq and last_active internal
state fields in the GIC. These fields don't correspond to state in
a real GIC and will be removed soon.
The changes to the ICSR are:
 * the VECTACTIVE field is documented as identical to the IPSR[8:0]
   field, so implement it that way
 * implement RETTOBASE via looking at the active state bits

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1438089748-5528-2-git-send-email-peter.maydell@linaro.org

9 years agopc_init1: pass parameters just with types
Tiejun Chen [Wed, 15 Jul 2015 05:37:42 +0000 (13:37 +0800)]
pc_init1: pass parameters just with types

Pass types to configure pc_init1().

Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
9 years agoi440fx: make types configurable at run-time
Michael S. Tsirkin [Wed, 15 Jul 2015 05:37:41 +0000 (13:37 +0800)]
i440fx: make types configurable at run-time

IGD passthrough wants to supply a different pci and
host devices, inheriting i440fx devices. Make types
configurable.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
9 years agoxen-hvm: Add trace to ioreq
Don Slutz [Thu, 30 Apr 2015 18:27:09 +0000 (14:27 -0400)]
xen-hvm: Add trace to ioreq

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Don Slutz <dslutz@verizon.com>
9 years agotarget-microblaze: Use setcond for pcmp*
Richard Henderson [Wed, 2 Sep 2015 21:46:01 +0000 (14:46 -0700)]
target-microblaze: Use setcond for pcmp*

Signed-off-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
9 years agotarget-cris: Use movcond and setcond
Richard Henderson [Wed, 2 Sep 2015 18:38:10 +0000 (11:38 -0700)]
target-cris: Use movcond and setcond

Signed-off-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
9 years agoMerge remote-tracking branch 'remotes/cohuck/tags/s390x-20150907' into staging
Peter Maydell [Mon, 7 Sep 2015 15:07:47 +0000 (16:07 +0100)]
Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20150907' into staging

s390x fixes and improvements:
- various bugfixes (css/event-facility)
- more efficient adapter interrupt routes setup
- gdb enhancement
- sclp got treated with a lot of remodelling/cleanup

# gpg: Signature made Mon 07 Sep 2015 15:42:43 BST using RSA key ID C6F02FAF
# gpg: Good signature from "Cornelia Huck <huckc@linux.vnet.ibm.com>"
# gpg:                 aka "Cornelia Huck <cornelia.huck@de.ibm.com>"

* remotes/cohuck/tags/s390x-20150907: (23 commits)
  s390/sclp: simplify calculation of rnmax
  s390/sclp: store the increment_size in the sclp device
  s390: unify allocation of initial memory
  s390: move memory calculation into the sclp device
  s390/sclp: ignore memory hotplug operations if it is disabled
  s390: disallow memory hotplug for the s390-virtio machine
  s390: no need to manually parse for slots and maxmem
  s390/sclp: move sclp_service_interrupt into the sclp device
  s390/sclp: move sclp_execute related functions into the SCLP class
  s390/sclp: introduce a root sclp device
  s390/sclp: temporarily fix unassignment/reassignment of memory subregions
  s390/sclp: replace sclp event types with proper defines
  s390/sclp: rework sclp event facility initialization + device realization
  sclp/s390: rework sclp cpu hotplug device notification
  s390x/gdb: support reading/writing of control registers
  s390x/kvm: make setting of in-kernel irq routes more efficient
  pc-bios/s390-ccw: rebuild image
  pc-bios/s390-ccw: Device detection in higher subchannel sets
  s390x/event-facility: fix location of receive mask
  s390x/css: start with cleared cstat/dstat
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agos390/sclp: simplify calculation of rnmax
David Hildenbrand [Mon, 1 Jun 2015 11:04:03 +0000 (13:04 +0200)]
s390/sclp: simplify calculation of rnmax

rnmax can be directly calculated using machine->maxram_size.

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
9 years agos390/sclp: store the increment_size in the sclp device
David Hildenbrand [Mon, 1 Jun 2015 11:03:23 +0000 (13:03 +0200)]
s390/sclp: store the increment_size in the sclp device

Let's calculate it once and reuse it.

Suggested-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
9 years agos390: unify allocation of initial memory
David Hildenbrand [Fri, 29 May 2015 13:01:55 +0000 (15:01 +0200)]
s390: unify allocation of initial memory

Now that the calculation of the initial memory is hidden in the sclp
device, we can unify the allocation of the initial memory.

The remaining ugly part is the reserved memory for the virtio queues,
but that can be cleaned up later.

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
9 years agos390: move memory calculation into the sclp device
David Hildenbrand [Fri, 29 May 2015 11:53:08 +0000 (13:53 +0200)]
s390: move memory calculation into the sclp device

The restrictions for memory calculation belong to the sclp device.

Let's move the calculation to that point, so we are able to unify it for
both s390 machines. The sclp device is the first device to be initialized.
It performs the calculation and safely stores it in the machine, where
other parts of the system can access an reuse it.

The memory hotplug device is now only created when it is really needed.

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
9 years agos390/sclp: ignore memory hotplug operations if it is disabled
David Hildenbrand [Fri, 29 May 2015 12:06:39 +0000 (14:06 +0200)]
s390/sclp: ignore memory hotplug operations if it is disabled

If no memory hotplug device was created, the sclp command facility is
not exposed (SCLP_FC_ASSIGN_ATTACH_READ_STOR). We therefore have no
memory hotplug and should correctly report SCLP_RC_INVALID_SCLP_COMMAND
if any such command is executed.

This gets rid of these ugly asserts that could have been triggered
for the s390-virtio machine.

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
9 years agos390: disallow memory hotplug for the s390-virtio machine
David Hildenbrand [Fri, 29 May 2015 11:22:12 +0000 (13:22 +0200)]
s390: disallow memory hotplug for the s390-virtio machine

That machine type doesn't currently support memory hotplug, so let's abort
if it is requested. Reason is, that the virtio queues are allocated for now
at the end of the initial ram - extending the ram is therefore not possible.

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
9 years agos390: no need to manually parse for slots and maxmem
David Hildenbrand [Fri, 29 May 2015 11:14:50 +0000 (13:14 +0200)]
s390: no need to manually parse for slots and maxmem

ram_slots and maxram_size has already been parsed and verified by
common code for us.

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
9 years agos390/sclp: move sclp_service_interrupt into the sclp device
David Hildenbrand [Wed, 13 May 2015 13:06:44 +0000 (15:06 +0200)]
s390/sclp: move sclp_service_interrupt into the sclp device

Let's make that function a method of the new sclp device, keeping
the wrapper for existing users.

We can now let go of get_event_facility().

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
9 years agos390/sclp: move sclp_execute related functions into the SCLP class
David Hildenbrand [Wed, 27 May 2015 08:04:56 +0000 (10:04 +0200)]
s390/sclp: move sclp_execute related functions into the SCLP class

Let's move the sclp_execute related functions into the SCLP class
and pass the device state as parameter, so we have easy access to
the SCLPDevice later on.

Reviewed-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>