pci: keep window properties up to date
cover pci: keep window properties up to date
w32/w64 properties that we report in QOM are currently static, but this is wrong:
guest firmware can select its own windows:
optimal placement for w64 is guest-dependent, further, for Q35, w32 is affected by the MCFG base and size.
This detects the actual window configuration used by guest and reports it in QOM.
Michael S. Tsirkin (6):
q35: make pci window address/size match guest cfg
range: add min/max operations on ranges
pci: add helper to retrieve the 64-bit range
range: add Range to typedefs
q35: use 64 bit window programmed by guest
piix: use 64 bit window programmed by guest
include/hw/pci/pci.h | 1 + include/qemu/range.h | 19 ++++++++++++++++++- include/qemu/typedefs.h | 1 + hw/pci-host/piix.c | 6 ++++++ hw/pci-host/q35.c | 16 ++++++++++++++++ hw/pci/pci.c | 43 +++++++++++++++++++++++++++++++++++++++++++ 6 files changed, 85 insertions(+), 1 deletion(-)
Signature · 1 line
MST
PATCH 1/6 q35: make pci window address/size match guest cfg
For Q35, MMCFG address and size are guest configurable. Update w32 property to make it behave accordingly.
--- hw/pci-host/q35.c | 10 ++++++++++ 1 file changed, 10 insertions(+)
hw/pci-host/q35.c +10 -0
@@ -214,6 +214,16 @@ static void mch_update_pciexbar(MCHPCIState *mch) } addr = pciexbar & addr_mask;pcie_host_mmcfg_update (pehb, enable, addr, length); +/* Leave enough space for the MCFG BAR */ +/* +* TODO: this matches current bios behaviour, but it's not a power of two, +* which means an MTRR can't cover it exactly. +*/ +if (enable) { + mch->pci_info.w32.begin = addr + length; + }else { + mch->pci_info.w32.begin =MCH_HOST_BRIDGE_PCIEXBAR_DEFAULT ; + } }/* PAM */
Signature · 1 line
MST
Re: [Qemu-devel] [PATCH 1/6] q35: make pci window address/size match guest cfg
Quoted 17 lines · On Wed, 4 Sep 2013 13:48:29 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
For Q35, MMCFG address and size are guest configurable. Update w32 property to make it behave accordingly.
Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 10 ++++++++++ 1 file changed, 10 insertions(+)hw/pci-host/q35.c +1
-0@@ -214,6 +214,16 @@ static void mch_update_pciexbar(MCHPCIState *mch) } addr = pciexbar & addr_mask;pcie_host_mmcfg_update (pehb, enable, addr, length); +/* Leave enough space for the MCFG BAR */
hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 227
Quoted 10 lines
hw/pci-host/q35.c +9
-0+/* +* TODO: this matches current bios behaviour, but it's not a power of two, +* which means an MTRR can't cover it exactly. +*/ +if (enable) { + mch->pci_info.w32.begin = addr + length; + }else { + mch->pci_info.w32.begin =MCH_HOST_BRIDGE_PCIEXBAR_DEFAULT ; + } }
I probably miss something but where is remapping in system address space? If there is none then, then updated w32 might mismatch actually/initially mapped alias.
Quoted 1 line
hw/pci-host/q35.c +0
-0/* PAM */
Re: [Qemu-devel] [PATCH 1/6] q35: make pci window address/size match guest cfg
Quoted 24 lines · Igor Mammedov
Quoted 21 lines · On Wed, 4 Sep 2013 13:48:29 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
For Q35, MMCFG address and size are guest configurable. Update w32 property to make it behave accordingly.
Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 10 ++++++++++ 1 file changed, 10 insertions(+)hw/pci-host/q35.c +5
-0@@ -214,6 +214,16 @@ static void mch_update_pciexbar(MCHPCIState *mch) } addr = pciexbar & addr_mask;pcie_host_mmcfg_update (pehb, enable, addr, length); +/* Leave enough space for the MCFG BAR */ +/* +* TODO: this matches current bios behaviour, but it's not a power of two, +* which means an MTRR can't cover it exactly. +*/
Quoted 10 lines
hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 227
Quoted 6 lines
hw/pci-host/q35.c +5
-0+if (enable) { + mch->pci_info.w32.begin = addr + length; + }else { + mch->pci_info.w32.begin =MCH_HOST_BRIDGE_PCIEXBAR_DEFAULT ; + } }I probably miss something but where is remapping in system address space? If there is none then, then updated w32 might mismatch actually/initially mapped alias.
Quoted 1 line
hw/pci-host/q35.c +0
-0/* PAM */
You mean mmcfg?
The re-mapping is in hw/pci/pcie_host.c
Re: [Qemu-devel] [PATCH 1/6] q35: make pci window address/size match guest cfg
Quoted 34 lines · On Tue, 10 Sep 2013 16:46:36 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 33 lines · Igor Mammedov
Quoted 17 lines · On Wed, 4 Sep 2013 13:48:29 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
For Q35, MMCFG address and size are guest configurable. Update w32 property to make it behave accordingly.
Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 10 ++++++++++ 1 file changed, 10 insertions(+)hw/pci-host/q35.c +1
-0@@ -214,6 +214,16 @@ static void mch_update_pciexbar(MCHPCIState *mch) } addr = pciexbar & addr_mask;pcie_host_mmcfg_update (pehb, enable, addr, length); +/* Leave enough space for the MCFG BAR */ hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 227
Quoted 10 lines
hw/pci-host/q35.c +9
-0+/* +* TODO: this matches current bios behaviour, but it's not a power of two, +* which means an MTRR can't cover it exactly. +*/ +if (enable) { + mch->pci_info.w32.begin = addr + length; + }else { + mch->pci_info.w32.begin =MCH_HOST_BRIDGE_PCIEXBAR_DEFAULT ; + } }I probably miss something but where is remapping in system address space? If there is none then, then updated w32 might mismatch actually/initially mapped alias.
Quoted 4 lines
Quoted 1 line
hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 229
Quoted 1 line
hw/pci-host/q35.c +0
-0/* PAM */ You mean mmcfg?
The re-mapping is inhw/pci/pcie_host.c
no, I mean 32-bit PCI hole
Re: [Qemu-devel] [PATCH 1/6] q35: make pci window address/size match guest cfg
Quoted 37 lines · Igor Mammedov
Quoted 34 lines · On Tue, 10 Sep 2013 16:46:36 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 33 lines · Igor Mammedov
Quoted 17 lines · On Wed, 4 Sep 2013 13:48:29 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
For Q35, MMCFG address and size are guest configurable. Update w32 property to make it behave accordingly.
Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 10 ++++++++++ 1 file changed, 10 insertions(+)hw/pci-host/q35.c +1
-0@@ -214,6 +214,16 @@ static void mch_update_pciexbar(MCHPCIState *mch) } addr = pciexbar & addr_mask;pcie_host_mmcfg_update (pehb, enable, addr, length); +/* Leave enough space for the MCFG BAR */ hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 227
Quoted 10 lines
hw/pci-host/q35.c +9
-0+/* +* TODO: this matches current bios behaviour, but it's not a power of two, +* which means an MTRR can't cover it exactly. +*/ +if (enable) { + mch->pci_info.w32.begin = addr + length; + }else { + mch->pci_info.w32.begin =MCH_HOST_BRIDGE_PCIEXBAR_DEFAULT ; + } }I probably miss something but where is remapping in system address space? If there is none then, then updated w32 might mismatch actually/initially mapped alias.
Quoted 5 lines
Quoted 4 lines
Quoted 1 line
hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 229
Quoted 1 line
hw/pci-host/q35.c +0
-0/* PAM */ You mean mmcfg?
The re-mapping is inhw/pci/pcie_host.cno, I mean 32-bit PCI hole
It works differently - see mch_init.
32-bit PCI hole is not remapped - instead mmcfg overlaps it and shadows a chunk of it.
Re: [Qemu-devel] [PATCH 1/6] q35: make pci window address/size match guest cfg
Quoted 40 lines · On Tue, 10 Sep 2013 16:53:58 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 39 lines · Igor Mammedov
Quoted 36 lines · On Tue, 10 Sep 2013 16:46:36 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 34 lines · Igor Mammedov
Quoted 17 lines · On Wed, 4 Sep 2013 13:48:29 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
For Q35, MMCFG address and size are guest configurable. Update w32 property to make it behave accordingly.
Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 10 ++++++++++ 1 file changed, 10 insertions(+)hw/pci-host/q35.c +1
-0@@ -214,6 +214,16 @@ static void mch_update_pciexbar(MCHPCIState *mch) } addr = pciexbar & addr_mask;pcie_host_mmcfg_update (pehb, enable, addr, length); +/* Leave enough space for the MCFG BAR */ hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 227
Quoted 10 lines
hw/pci-host/q35.c +9
-0+/* +* TODO: this matches current bios behaviour, but it's not a power of two, +* which means an MTRR can't cover it exactly. +*/ +if (enable) { + mch->pci_info.w32.begin = addr + length; + }else { + mch->pci_info.w32.begin =MCH_HOST_BRIDGE_PCIEXBAR_DEFAULT ; + } }I probably miss something but where is remapping in system address space? If there is none then, then updated w32 might mismatch actually/initially mapped alias.
Quoted 1 line
hw/pci-host/q35.c +0
-0/* PAM */
Quoted 8 lines
Quoted 3 lines
Quoted 2 lines
You mean mmcfg?
The re-mapping is inhw/pci/pcie_host.cno, I mean 32-bit PCI hole
It works differently - see
mch_init.
32-bit PCI hole is not remapped - instead mmcfg overlaps it and shadows a chunk of it.
it probably won't cause harm, but actual pci_hole alias doesn't match its window reported via properties anymore (it's just inconsistent).
Re: [Qemu-devel] [PATCH 1/6] q35: make pci window address/size match guest cfg
Quoted 45 lines · Igor Mammedov
Quoted 42 lines · On Tue, 10 Sep 2013 16:53:58 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 41 lines · Igor Mammedov
Quoted 38 lines · On Tue, 10 Sep 2013 16:46:36 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 24 lines · Igor Mammedov
Quoted 21 lines · On Wed, 4 Sep 2013 13:48:29 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
For Q35, MMCFG address and size are guest configurable. Update w32 property to make it behave accordingly.
Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 10 ++++++++++ 1 file changed, 10 insertions(+)hw/pci-host/q35.c +5
-0@@ -214,6 +214,16 @@ static void mch_update_pciexbar(MCHPCIState *mch) } addr = pciexbar & addr_mask;pcie_host_mmcfg_update (pehb, enable, addr, length); +/* Leave enough space for the MCFG BAR */ +/* +* TODO: this matches current bios behaviour, but it's not a power of two, +* which means an MTRR can't cover it exactly. +*/ Quoted 10 lines
hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 227
Quoted 6 lines
hw/pci-host/q35.c +5
-0+if (enable) { + mch->pci_info.w32.begin = addr + length; + }else { + mch->pci_info.w32.begin =MCH_HOST_BRIDGE_PCIEXBAR_DEFAULT ; + } }I probably miss something but where is remapping in system address space? If there is none then, then updated w32 might mismatch actually/initially mapped alias.
Quoted 1 line
hw/pci-host/q35.c +0
-0/* PAM */ You mean mmcfg?
The re-mapping is inhw/pci/pcie_host.c
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
no, I mean 32-bit PCI hole
It works differently - see
mch_init.
32-bit PCI hole is not remapped - instead mmcfg overlaps it and shadows a chunk of it.it probably won't cause harm, but actual
pci_holealias doesn't match its window reported via properties anymore (it's just inconsistent).
That's exactly what happens at the moment too:
for q35 pci hole has a chunk that is not reported in w32.
Do you imply that w32 should just report the PCI hole, and make users (e.g. acpi) take mmcfg out of that?
Signature · 1 line
MST
Re: [Qemu-devel] [PATCH 1/6] q35: make pci window address/size match guest cfg
Quoted 52 lines · On Tue, 10 Sep 2013 17:16:24 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 51 lines · Igor Mammedov
Quoted 48 lines · On Tue, 10 Sep 2013 16:53:58 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 37 lines · Igor Mammedov
Quoted 34 lines · On Tue, 10 Sep 2013 16:46:36 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 33 lines · Igor Mammedov
Quoted 17 lines · On Wed, 4 Sep 2013 13:48:29 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
For Q35, MMCFG address and size are guest configurable. Update w32 property to make it behave accordingly.
Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 10 ++++++++++ 1 file changed, 10 insertions(+)hw/pci-host/q35.c +1
-0@@ -214,6 +214,16 @@ static void mch_update_pciexbar(MCHPCIState *mch) } addr = pciexbar & addr_mask;pcie_host_mmcfg_update (pehb, enable, addr, length); +/* Leave enough space for the MCFG BAR */ hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 227
Quoted 10 lines
hw/pci-host/q35.c +9
-0+/* +* TODO: this matches current bios behaviour, but it's not a power of two, +* which means an MTRR can't cover it exactly. +*/ +if (enable) { + mch->pci_info.w32.begin = addr + length; + }else { + mch->pci_info.w32.begin =MCH_HOST_BRIDGE_PCIEXBAR_DEFAULT ; + } }I probably miss something but where is remapping in system address space? If there is none then, then updated w32 might mismatch actually/initially mapped alias.
Quoted 5 lines
Quoted 4 lines
Quoted 1 line
hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 228
Quoted 1 line
hw/pci-host/q35.c +0
-0/* PAM */ You mean mmcfg?
The re-mapping is inhw/pci/pcie_host.cno, I mean 32-bit PCI hole
It works differently - see
mch_init.
32-bit PCI hole is not remapped - instead mmcfg overlaps it and shadows a chunk of it.
Quoted 8 lines
Quoted 2 lines
it probably won't cause harm, but actual
pci_holealias doesn't match its window reported via properties anymore (it's just inconsistent).That's exactly what happens at the moment too:
for q35 pci hole has a chunk that is not reported in w32.Do you imply that w32 should just report the PCI hole, and make users (e.g. acpi) take mmcfg out of that?
initial pci_hole alias in system AS doesn't include mmcfg, it's mapped right after mmcfg and w32 range matches alias exactly. But after mmcfg update that might be not true. I suggest to move/remap alias to match new w32, to be consistent.
Quoted 2 lines
Re: [Qemu-devel] [PATCH 1/6] q35: make pci window address/size match guest cfg
Quoted 58 lines · Igor Mammedov
Quoted 55 lines · On Tue, 10 Sep 2013 17:16:24 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 53 lines · Igor Mammedov
Quoted 40 lines · On Tue, 10 Sep 2013 16:53:58 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 39 lines · Igor Mammedov
Quoted 36 lines · On Tue, 10 Sep 2013 16:46:36 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 34 lines · Igor Mammedov
Quoted 17 lines · On Wed, 4 Sep 2013 13:48:29 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
For Q35, MMCFG address and size are guest configurable. Update w32 property to make it behave accordingly.
Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 10 ++++++++++ 1 file changed, 10 insertions(+)hw/pci-host/q35.c +1
-0@@ -214,6 +214,16 @@ static void mch_update_pciexbar(MCHPCIState *mch) } addr = pciexbar & addr_mask;pcie_host_mmcfg_update (pehb, enable, addr, length); +/* Leave enough space for the MCFG BAR */ hw/pci-host/q35.c static void mch_update_pciexbar(MCHPCIState *mch) · line 227
Quoted 10 lines
hw/pci-host/q35.c +9
-0+/* +* TODO: this matches current bios behaviour, but it's not a power of two, +* which means an MTRR can't cover it exactly. +*/ +if (enable) { + mch->pci_info.w32.begin = addr + length; + }else { + mch->pci_info.w32.begin =MCH_HOST_BRIDGE_PCIEXBAR_DEFAULT ; + } }I probably miss something but where is remapping in system address space? If there is none then, then updated w32 might mismatch actually/initially mapped alias.
Quoted 1 line
hw/pci-host/q35.c +0
-0/* PAM */ Quoted 8 lines
Quoted 3 lines
Quoted 2 lines
You mean mmcfg?
The re-mapping is inhw/pci/pcie_host.cno, I mean 32-bit PCI hole
It works differently - see
mch_init.
32-bit PCI hole is not remapped - instead mmcfg overlaps it and shadows a chunk of it.it probably won't cause harm, but actual
pci_holealias doesn't match its window reported via properties anymore (it's just inconsistent).
Quoted 8 lines
Quoted 5 lines
That's exactly what happens at the moment too:
for q35 pci hole has a chunk that is not reported in w32.Do you imply that w32 should just report the PCI hole, and make users (e.g. acpi) take mmcfg out of that?
initial
pci_holealias in system AS doesn't include mmcfg, it's mapped right after mmcfg and w32 range matches alias exactly.
That would be a bug (unrelated to this patch series). But I don't see this code: looking at mch_init,
memory_region_add_subregion(mch->system_memory, mch->below_4g_mem_size,
&mch->pci_hole);so we have pci hole right after memory, and overlapping mmcfg, which makes sense to me.
Quoted 7 lines
But after mmcfg update that might be not true. I suggest to move/remap alias to match new w32, to be consistent.
Quoted 2 lines
PATCH 2/6 range: add Range to typedefs
In reply to Michael S. Tsirkinwill help simplify header dependencies.
--- include/qemu/range.h | 2 +- include/qemu/typedefs.h | 1 + 2 files changed, 2 insertions(+), 1 deletion(-)
include/qemu/range.h +1 -1
@@ -2,6 +2,7 @@#define QEMU_RANGE_H #include <inttypes.h> +#include <qemu/typedefs.h> /* * Operations on 64 bit address ranges.
@@ -15,7 +16,6 @@ struct Range {uint64_t begin;/* First byte of the range, or 0 if empty. */ uint64_t end;/* 1 + the last byte. 0 if range empty or ends at ~0x0LL. */ }; -typedef struct Range Range ;/* Get last byte of a range from offset + length. * Undefined for ranges that wrap around 0. */
include/qemu/typedefs.h +1 -0
@@ -65,5 +65,6 @@ typedef struct QEMUSGList QEMUSGList;typedef struct SHPCDevice SHPCDevice ;typedef struct FWCfgState FWCfgState ;typedef struct PcGuestInfo PcGuestInfo ; +typedef struct Range Range ;#endif /* QEMU_TYPEDEFS_H */
Signature · 1 line
MST
PATCH 3/6 range: add min/max operations on ranges
In reply to Michael S. Tsirkin--- include/qemu/range.h | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)
include/qemu/range.h +17 -0
@@ -17,6 +17,23 @@ struct Range {uint64_t end;/* 1 + the last byte. 0 if range empty or ends at ~0x0LL. */ }; +static inline void range_extend (Range *range,Range *extend_by) +{ +if (!extend_by->begin && !extend_by->end) { +return ; + } +if (!range->begin && !range->end) { + *range = *extend_by; +return ; + } +if (range->begin > extend_by->begin) { + range->begin = extend_by->begin; + } +if (range->end -1 < extend_by->end -1 ) { + range->end = extend_by->end; + } +} +/* Get last byte of a range from offset + length. * Undefined for ranges that wrap around 0. */ static inline uint64_t range_get_last (uint64_t offset,uint64_t len)
Signature · 1 line
MST
Re: [Qemu-devel] [PATCH 3/6] range: add min/max operations on ranges
Quoted 10 lines · On Wed, 4 Sep 2013 13:48:35 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Signed-off-by Michael S. Tsirkin <[email protected]>--- include/qemu/range.h | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)include/qemu/range.h +0
-0
include/qemu/range.h struct Range { · line 20
Quoted 4 lines
include/qemu/range.h +1
-0@@ -17,6 +17,23 @@ struct Range {uint64_t end;/* 1 + the last byte. 0 if range empty or ends at ~0x0LL. */ }; +static inline void range_extend (Range *range,Range *extend_by)
doc comment what it does pls.
include/qemu/range.h struct Range { · line 32
Quoted 12 lines
include/qemu/range.h +12
-0+{ +if (!extend_by->begin && !extend_by->end) { +return ; + } +if (!range->begin && !range->end) { + *range = *extend_by; +return ; + } +if (range->begin > extend_by->begin) { + range->begin = extend_by->begin; + } +if (range->end -1 < extend_by->end -1 ) {
(foo)->end could be 0 at this point leading to overflow when subtracted, is it intended to be so?
Quoted 7 lines
include/qemu/range.h +4
-0+ range->end = extend_by->end; + } +} +/* Get last byte of a range from offset + length. * Undefined for ranges that wrap around 0. */ static inline uint64_t range_get_last (uint64_t offset,uint64_t len)
Re: [Qemu-devel] [PATCH 3/6] range: add min/max operations on ranges
Quoted 31 lines · Igor Mammedov
Quoted 10 lines · On Wed, 4 Sep 2013 13:48:35 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Signed-off-by Michael S. Tsirkin <[email protected]>--- include/qemu/range.h | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)include/qemu/range.h +0
-0include/qemu/range.h struct Range { · line 20
Quoted 4 lines
include/qemu/range.h +1
-0@@ -17,6 +17,23 @@ struct Range {uint64_t end;/* 1 + the last byte. 0 if range empty or ends at ~0x0LL. */ }; +static inline void range_extend (Range *range,Range *extend_by)doc comment what it does pls.
include/qemu/range.h struct Range { · line 32
Quoted 12 lines
include/qemu/range.h +12
-0+{ +if (!extend_by->begin && !extend_by->end) { +return ; + } +if (!range->begin && !range->end) { + *range = *extend_by; +return ; + } +if (range->begin > extend_by->begin) { + range->begin = extend_by->begin; + } +if (range->end -1 < extend_by->end -1 ) {
Quoted 2 lines
(foo)->end could be 0 at this point leading to overflow when subtracted, is it intended to be so?
Absolutely - as the comment near this field definition states:
0 means region ends at ~0x0LL.
Quoted 8 lines
Quoted 7 lines
include/qemu/range.h +4
-0+ range->end = extend_by->end; + } +} +/* Get last byte of a range from offset + length. * Undefined for ranges that wrap around 0. */ static inline uint64_t range_get_last (uint64_t offset,uint64_t len)
PATCH 4/6 pci: add helper to retrieve the 64-bit range
In reply to Michael S. Tsirkin--- include/hw/pci/pci.h | 1 + hw/pci/pci.c | 43 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 44 insertions(+)
include/hw/pci/pci.h +1 -0
@@ -397,6 +397,7 @@ const char *pci_root_bus_path(PCIDevice *dev);PCIDevice *pci_find_device (PCIBus *bus,int bus_num,uint8_t devfn);int pci_qdev_find_device (const char *id,PCIDevice **pdev);PCIBus *pci_get_bus_devfn (int *devfnp,PCIBus *root,const char *devaddr); +void pci_bus_get_w64_range (PCIBus *bus,Range *range);int pci_parse_devaddr (const char *addr,int *domp,int *busp,unsigned int *slotp,unsigned int *funcp);
hw/pci/pci.c +43 -0
49 lines not on this page · raw message
Signature · 1 line
MST
Re: [Qemu-devel] [PATCH 4/6] pci: add helper to retrieve the 64-bit range
Quoted 29 lines · On Wed, 4 Sep 2013 13:48:37 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Signed-off-by Michael S. Tsirkin <[email protected]>--- include/hw/pci/pci.h | 1 + hw/pci/pci.c | 43 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 44 insertions(+)include/hw/pci/pci.h +1
-0@@ -397,6 +397,7 @@ const char *pci_root_bus_path(PCIDevice *dev);PCIDevice *pci_find_device (PCIBus *bus,int bus_num,uint8_t devfn);int pci_qdev_find_device (const char *id,PCIDevice **pdev);PCIBus *pci_get_bus_devfn (int *devfnp,PCIBus *root,const char *devaddr); +void pci_bus_get_w64_range (PCIBus *bus,Range *range);int pci_parse_devaddr (const char *addr,int *domp,int *busp,unsigned int *slotp,unsigned int *funcp);hw/pci/pci.c +3
-0@@ -2257,6 +2257,49 @@ void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, void *opaque) bus->iommu_opaque = opaque; } +static void pci_dev_get_w64 (PCIBus *b,PCIDevice *dev,void *opaque) +{ +Range *range = opaque;
hw/pci/pci.c void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, void *opaque) · line 2272
Quoted 10 lines
hw/pci/pci.c +10
-0+PCIDeviceClass *pc =PCI_DEVICE_GET_CLASS (dev); +uint16_t cmd =pci_get_word (dev->config +PCI_COMMAND ); +int r; + +if (!(cmd &PCI_COMMAND_MEMORY )) { +return ; + } + +if (pc->is_bridge) { +pcibus_t base =pci_bridge_get_base (dev,PCI_BASE_ADDRESS_MEM_PREFETCH );
what guaranties that 'base' won't be below 4Gb and be above '4Gb + above_4g_mem_size'?
Quoted 33 lines
hw/pci/pci.c +30
-0+pcibus_t limit =pci_bridge_get_limit (dev,PCI_BASE_ADDRESS_MEM_PREFETCH ); + +if (limit >= base) { +Range pref_range; + pref_range.begin = base; + pref_range.end = limit +1 ; +range_extend (range, &pref_range); + } + } +for (r =0 ; r <PCI_NUM_REGIONS ; ++r) { +PCIIORegion *region = &dev->io_regions[r]; +Range region_range; + +if (!region->size || + (region->type &PCI_BASE_ADDRESS_SPACE_IO ) || + !(region->type &PCI_BASE_ADDRESS_MEM_TYPE_64 )) { +continue ; + } + region_range.begin =pci_get_quad (dev->config +pci_bar (dev, r)); + region_range.end = region_range.begin + region->size; +range_extend (range, ®ion_range); + } +} + +void pci_bus_get_w64_range (PCIBus *bus,Range *range) +{ + range->begin = range->end =0 ; +pci_for_each_device_under_bus (bus, pci_dev_get_w64, range); +} +static const TypeInfo pci_device_type_info = { .name =TYPE_PCI_DEVICE , .parent =TYPE_DEVICE ,
Re: [Qemu-devel] [PATCH 4/6] pci: add helper to retrieve the 64-bit range
Quoted 29 lines · Igor Mammedov
Quoted 26 lines · On Wed, 4 Sep 2013 13:48:37 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Signed-off-by Michael S. Tsirkin <[email protected]>--- include/hw/pci/pci.h | 1 + hw/pci/pci.c | 43 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 44 insertions(+)include/hw/pci/pci.h +1
-0@@ -397,6 +397,7 @@ const char *pci_root_bus_path(PCIDevice *dev);PCIDevice *pci_find_device (PCIBus *bus,int bus_num,uint8_t devfn);int pci_qdev_find_device (const char *id,PCIDevice **pdev);PCIBus *pci_get_bus_devfn (int *devfnp,PCIBus *root,const char *devaddr); +void pci_bus_get_w64_range (PCIBus *bus,Range *range);int pci_parse_devaddr (const char *addr,int *domp,int *busp,unsigned int *slotp,unsigned int *funcp);hw/pci/pci.c +0
-0@@ -2257,6 +2257,49 @@ void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, void *opaque) bus->iommu_opaque = opaque; }
Quoted 14 lines
hw/pci/pci.c void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, void *opaque) · line 2272
Quoted 13 lines
hw/pci/pci.c +13
-0+static void pci_dev_get_w64 (PCIBus *b,PCIDevice *dev,void *opaque) +{ +Range *range = opaque; +PCIDeviceClass *pc =PCI_DEVICE_GET_CLASS (dev); +uint16_t cmd =pci_get_word (dev->config +PCI_COMMAND ); +int r; + +if (!(cmd &PCI_COMMAND_MEMORY )) { +return ; + } + +if (pc->is_bridge) { +pcibus_t base =pci_bridge_get_base (dev,PCI_BASE_ADDRESS_MEM_PREFETCH );what guaranties that 'base' won't be below 4Gb
Hmm, this needs some thought.
Quoted 1 line
and be above '4Gb +
above_4g_mem_size'?
This one is harmless - guest will only hurt itself if it misconfigures the bridge like this.
Quoted 33 lines
Quoted 33 lines
hw/pci/pci.c +30
-0+pcibus_t limit =pci_bridge_get_limit (dev,PCI_BASE_ADDRESS_MEM_PREFETCH ); + +if (limit >= base) { +Range pref_range; + pref_range.begin = base; + pref_range.end = limit +1 ; +range_extend (range, &pref_range); + } + } +for (r =0 ; r <PCI_NUM_REGIONS ; ++r) { +PCIIORegion *region = &dev->io_regions[r]; +Range region_range; + +if (!region->size || + (region->type &PCI_BASE_ADDRESS_SPACE_IO ) || + !(region->type &PCI_BASE_ADDRESS_MEM_TYPE_64 )) { +continue ; + } + region_range.begin =pci_get_quad (dev->config +pci_bar (dev, r)); + region_range.end = region_range.begin + region->size; +range_extend (range, ®ion_range); + } +} + +void pci_bus_get_w64_range (PCIBus *bus,Range *range) +{ + range->begin = range->end =0 ; +pci_for_each_device_under_bus (bus, pci_dev_get_w64, range); +} +static const TypeInfo pci_device_type_info = { .name =TYPE_PCI_DEVICE , .parent =TYPE_DEVICE ,
Re: [Qemu-devel] [PATCH 4/6] pci: add helper to retrieve the 64-bit range
Quoted 45 lines · On Tue, 10 Sep 2013 16:50:41 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 29 lines · Igor Mammedov
Quoted 26 lines · On Wed, 4 Sep 2013 13:48:37 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Signed-off-by Michael S. Tsirkin <[email protected]>--- include/hw/pci/pci.h | 1 + hw/pci/pci.c | 43 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 44 insertions(+)include/hw/pci/pci.h +1
-0@@ -397,6 +397,7 @@ const char *pci_root_bus_path(PCIDevice *dev);PCIDevice *pci_find_device (PCIBus *bus,int bus_num,uint8_t devfn);int pci_qdev_find_device (const char *id,PCIDevice **pdev);PCIBus *pci_get_bus_devfn (int *devfnp,PCIBus *root,const char *devaddr); +void pci_bus_get_w64_range (PCIBus *bus,Range *range);int pci_parse_devaddr (const char *addr,int *domp,int *busp,unsigned int *slotp,unsigned int *funcp);hw/pci/pci.c +0
-0@@ -2257,6 +2257,49 @@ void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, void *opaque) bus->iommu_opaque = opaque; }Quoted 14 lines
hw/pci/pci.c void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, void *opaque) · line 2272
Quoted 13 lines
hw/pci/pci.c +13
-0+static void pci_dev_get_w64 (PCIBus *b,PCIDevice *dev,void *opaque) +{ +Range *range = opaque; +PCIDeviceClass *pc =PCI_DEVICE_GET_CLASS (dev); +uint16_t cmd =pci_get_word (dev->config +PCI_COMMAND ); +int r; + +if (!(cmd &PCI_COMMAND_MEMORY )) { +return ; + } + +if (pc->is_bridge) { +pcibus_t base =pci_bridge_get_base (dev,PCI_BASE_ADDRESS_MEM_PREFETCH );what guaranties that 'base' won't be below 4Gb
Quoted 6 lines
Hmm, this needs some thought.
Quoted 1 line
and be above '4Gb +
above_4g_mem_size'?This one is harmless - guest will only hurt itself if it misconfigures the bridge like this.
start of w64 is hardcoded in QEMU and Seabios as (4Gb+above_4gb_ram_size), so maybe there is no need to update/modify it at all?
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
hw/pci/pci.c +30
-0+pcibus_t limit =pci_bridge_get_limit (dev,PCI_BASE_ADDRESS_MEM_PREFETCH ); + +if (limit >= base) { +Range pref_range; + pref_range.begin = base; + pref_range.end = limit +1 ; +range_extend (range, &pref_range); + } + } +for (r =0 ; r <PCI_NUM_REGIONS ; ++r) { +PCIIORegion *region = &dev->io_regions[r]; +Range region_range; + +if (!region->size || + (region->type &PCI_BASE_ADDRESS_SPACE_IO ) || + !(region->type &PCI_BASE_ADDRESS_MEM_TYPE_64 )) { +continue ; + } + region_range.begin =pci_get_quad (dev->config +pci_bar (dev, r)); + region_range.end = region_range.begin + region->size; +range_extend (range, ®ion_range); + } +} + +void pci_bus_get_w64_range (PCIBus *bus,Range *range) +{ + range->begin = range->end =0 ; +pci_for_each_device_under_bus (bus, pci_dev_get_w64, range); +} +static const TypeInfo pci_device_type_info = { .name =TYPE_PCI_DEVICE , .parent =TYPE_DEVICE ,
Re: [Qemu-devel] [PATCH 4/6] pci: add helper to retrieve the 64-bit range
Quoted 50 lines · Igor Mammedov
Quoted 47 lines · On Tue, 10 Sep 2013 16:50:41 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Quoted 29 lines · Igor Mammedov
Quoted 26 lines · On Wed, 4 Sep 2013 13:48:37 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Signed-off-by Michael S. Tsirkin <[email protected]>--- include/hw/pci/pci.h | 1 + hw/pci/pci.c | 43 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 44 insertions(+)include/hw/pci/pci.h +1
-0@@ -397,6 +397,7 @@ const char *pci_root_bus_path(PCIDevice *dev);PCIDevice *pci_find_device (PCIBus *bus,int bus_num,uint8_t devfn);int pci_qdev_find_device (const char *id,PCIDevice **pdev);PCIBus *pci_get_bus_devfn (int *devfnp,PCIBus *root,const char *devaddr); +void pci_bus_get_w64_range (PCIBus *bus,Range *range);int pci_parse_devaddr (const char *addr,int *domp,int *busp,unsigned int *slotp,unsigned int *funcp);hw/pci/pci.c +0
-0@@ -2257,6 +2257,49 @@ void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, void *opaque) bus->iommu_opaque = opaque; }Quoted 14 lines
hw/pci/pci.c void pci_setup_iommu(PCIBus *bus, PCIIOMMUFunc fn, void *opaque) · line 2272
Quoted 13 lines
hw/pci/pci.c +13
-0+static void pci_dev_get_w64 (PCIBus *b,PCIDevice *dev,void *opaque) +{ +Range *range = opaque; +PCIDeviceClass *pc =PCI_DEVICE_GET_CLASS (dev); +uint16_t cmd =pci_get_word (dev->config +PCI_COMMAND ); +int r; + +if (!(cmd &PCI_COMMAND_MEMORY )) { +return ; + } + +if (pc->is_bridge) { +pcibus_t base =pci_bridge_get_base (dev,PCI_BASE_ADDRESS_MEM_PREFETCH );what guaranties that 'base' won't be below 4Gb
Hmm, this needs some thought.
Quoted 8 lines
Quoted 4 lines
Quoted 1 line
and be above '4Gb +
above_4g_mem_size'?This one is harmless - guest will only hurt itself if it misconfigures the bridge like this.
start of w64 is hardcoded in QEMU and Seabios as (4Gb+above_4gb_ram_size), so maybe there is no need to update/modify it at all?
Well the point is to have QEMU obey whatever bios does. seabios hardcodes the start of window but some other firmware might not.
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
hw/pci/pci.c +30
-0+pcibus_t limit =pci_bridge_get_limit (dev,PCI_BASE_ADDRESS_MEM_PREFETCH ); + +if (limit >= base) { +Range pref_range; + pref_range.begin = base; + pref_range.end = limit +1 ; +range_extend (range, &pref_range); + } + } +for (r =0 ; r <PCI_NUM_REGIONS ; ++r) { +PCIIORegion *region = &dev->io_regions[r]; +Range region_range; + +if (!region->size || + (region->type &PCI_BASE_ADDRESS_SPACE_IO ) || + !(region->type &PCI_BASE_ADDRESS_MEM_TYPE_64 )) { +continue ; + } + region_range.begin =pci_get_quad (dev->config +pci_bar (dev, r)); + region_range.end = region_range.begin + region->size; +range_extend (range, ®ion_range); + } +} + +void pci_bus_get_w64_range (PCIBus *bus,Range *range) +{ + range->begin = range->end =0 ; +pci_for_each_device_under_bus (bus, pci_dev_get_w64, range); +} +static const TypeInfo pci_device_type_info = { .name =TYPE_PCI_DEVICE , .parent =TYPE_DEVICE ,
PATCH 5/6 q35: use 64 bit window programmed by guest
In reply to Michael S. TsirkinDetect the 64 bit window programmed by firmware
and configure properties accordingly.
--- hw/pci-host/q35.c | 6 ++++++ 1 file changed, 6 insertions(+)
hw/pci-host/q35.c +6 -0
@@ -90,6 +90,9 @@ static void q35_host_get_pci_hole64_start(Object *obj, Visitor *v, Error **errp) {Q35PCIHost *s =Q35_HOST_DEVICE (obj); +PCIHostState *h =PCI_HOST_BRIDGE (obj); + +pci_bus_get_w64_range (h->bus, &s->mch.pci_info.w64);visit_type_uint64 (v, &s->mch.pci_info.w64.begin, name, errp); }
@@ -99,6 +102,9 @@ static void q35_host_get_pci_hole64_end(Object *obj, Visitor *v, Error **errp) {Q35PCIHost *s =Q35_HOST_DEVICE (obj); +PCIHostState *h =PCI_HOST_BRIDGE (obj); + +pci_bus_get_w64_range (h->bus, &s->mch.pci_info.w64);visit_type_uint64 (v, &s->mch.pci_info.w64.end, name, errp); }
Signature · 1 line
MST
Re: [Qemu-devel] [PATCH 5/6] q35: use 64 bit window programmed by guest
Quoted 13 lines · On Wed, 4 Sep 2013 13:48:40 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Detect the 64 bit window programmed by firmware
and configure properties accordingly.Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 6 ++++++ 1 file changed, 6 insertions(+)hw/pci-host/q35.c +0
-0
hw/pci-host/q35.c static void q35_host_get_pci_hole64_start(Object *obj, Visitor *v, · line 95
Quoted 6 lines
hw/pci-host/q35.c +3
-0@@ -90,6 +90,9 @@ static void q35_host_get_pci_hole64_start(Object *obj, Visitor *v, Error **errp) {Q35PCIHost *s =Q35_HOST_DEVICE (obj); +PCIHostState *h =PCI_HOST_BRIDGE (obj); + +pci_bus_get_w64_range (h->bus, &s->mch.pci_info.w64);
Shouldn't be it done at the time when BIOS initializes BAR, to avoid inconsistent state in between write and read.
Also missing remapping of existing 64-bit PCI hole alias to a new range.
hw/pci-host/q35.c static void q35_host_get_pci_hole64_end(Object *obj, Visitor *v, · line 107
Quoted 10 lines
hw/pci-host/q35.c +3
-0visit_type_uint64 (v, &s->mch.pci_info.w64.begin, name, errp); }@@ -99,6 +102,9 @@ static void q35_host_get_pci_hole64_end(Object *obj, Visitor *v, Error **errp) {Q35PCIHost *s =Q35_HOST_DEVICE (obj); +PCIHostState *h =PCI_HOST_BRIDGE (obj); + +pci_bus_get_w64_range (h->bus, &s->mch.pci_info.w64);
ditto
Quoted 3 lines
hw/pci-host/q35.c +0
-0visit_type_uint64 (v, &s->mch.pci_info.w64.end, name, errp); }
Re: [Qemu-devel] [PATCH 5/6] q35: use 64 bit window programmed by guest
Quoted 11 lines · Igor Mammedov
Quoted 8 lines · On Wed, 4 Sep 2013 13:48:40 +0300 "Michael S. Tsirkin" <[email protected]> wrote:
Detect the 64 bit window programmed by firmware
and configure properties accordingly.Signed-off-by Michael S. Tsirkin <[email protected]>--- hw/pci-host/q35.c | 6 ++++++ 1 file changed, 6 insertions(+)
Quoted 13 lines
hw/pci-host/q35.c static void q35_host_get_pci_hole64_start(Object *obj, Visitor *v, · line 95
Quoted 11 lines
hw/pci-host/q35.c +3
-0@@ -90,6 +90,9 @@ static void q35_host_get_pci_hole64_start(Object *obj, Visitor *v, Error **errp) {Q35PCIHost *s =Q35_HOST_DEVICE (obj); +PCIHostState *h =PCI_HOST_BRIDGE (obj); + +pci_bus_get_w64_range (h->bus, &s->mch.pci_info.w64);Shouldn't be it done at the time when BIOS initializes BAR, to avoid inconsistent state in between write and read.
What's inconsistent with what?
We just report the current state.
If you query it during initialization, you get what has been programmed so far.
Quoted 1 line
Also missing remapping of existing 64-bit PCI hole alias to a new range.
We shouldn't remap the region - this is not how hardware works. This property merely reports what BIOS decided to use.
Quoted 15 lines
hw/pci-host/q35.c static void q35_host_get_pci_hole64_end(Object *obj, Visitor *v, · line 107
Quoted 10 lines
hw/pci-host/q35.c +3
-0visit_type_uint64 (v, &s->mch.pci_info.w64.begin, name, errp); }@@ -99,6 +102,9 @@ static void q35_host_get_pci_hole64_end(Object *obj, Visitor *v, Error **errp) {Q35PCIHost *s =Q35_HOST_DEVICE (obj); +PCIHostState *h =PCI_HOST_BRIDGE (obj); + +pci_bus_get_w64_range (h->bus, &s->mch.pci_info.w64);ditto
Quoted 3 lines
hw/pci-host/q35.c +0
-0visit_type_uint64 (v, &s->mch.pci_info.w64.end, name, errp); }
PATCH 6/6 piix: use 64 bit window programmed by guest
In reply to Michael S. TsirkinDetect the 64 bit window programmed by firmware
and configure properties accordingly.
--- hw/pci-host/piix.c | 6 ++++++ 1 file changed, 6 insertions(+)
hw/pci-host/piix.c +6 -0
@@ -236,6 +236,9 @@ static void i440fx_pcihost_get_pci_hole64_start(Object *obj, Visitor *v, Error **errp) {I440FXState *s =I440FX_PCI_HOST_BRIDGE (obj); +PCIHostState *h =PCI_HOST_BRIDGE (obj); + +pci_bus_get_w64_range (h->bus, &s->pci_info.w64);visit_type_uint64 (v, &s->pci_info.w64.begin, name, errp); }
@@ -245,6 +248,9 @@ static void i440fx_pcihost_get_pci_hole64_end(Object *obj, Visitor *v, Error **errp) {I440FXState *s =I440FX_PCI_HOST_BRIDGE (obj); +PCIHostState *h =PCI_HOST_BRIDGE (obj); + +pci_bus_get_w64_range (h->bus, &s->pci_info.w64);visit_type_uint64 (v, &s->pci_info.w64.end, name, errp); }
Signature · 1 line
MST