The qemu part of this patch can be found in https://lists.gnu.org/archive/html/qemu-devel/2018-09/msg01988.html
pxb-pcie device in qemu uses only one pci domain(0) so far, which means there are at most 256 busses. However, pcie topology requires one bus per device, which will use up the busses soon if there are more pcie devices.
To solve the problem, QEMU can put the pxb-pcie in separate pci domain so we can get more devices. As QEMU relies on seabios to allocate space for its mcfg table, we must modify the seabios to configure mcfg_base and mcfg_size for qemu.
Since pxb-pcie may only uses a subset of all 256 busses in a domain, we let qemu pass its desired mcfg_size as a hidden bar, seabios just decides the mcfg_base for it.
Compared with previous design, mmconfig is not put above 4g anymore, but inside [0x80000000, 0xb0000000), and leaves the major part of configuration to guest OS, which makes the whole part of code much simpler.
Changelog: v3 <- v2 - Refactor design so that seabios only does minimal mcfg configure, and leaves everything else to guest os. This makes the code much simpler. - Does not put mcfg over 4g anymore, but inside [0x80000000, 0xb0000000) Over 4g placement can be expected in future version.
v2 <- v1 - Fix bugs in filtering domains when traversing pci devices - Reformat some hardcoded codes, such as probing the pci device in pci_setup
Zihan Yang (1): pciinit: setup mcfg for pxb-pcie to support multiple pci domains
src/fw/dev-q35.h | 7 +++++++ src/fw/pciinit.c | 32 ++++++++++++++++++++++++++++++++ src/hw/pci_ids.h | 1 + 3 files changed, 40 insertions(+)
To support multiple pci domains of pxb-pcie device in qemu, we need to setup mcfg range in seabios. We use [0x80000000, 0xb0000000) to hold new domain mcfg table for now, and we need to retrieve the desired mcfg size of each pxb-pcie from a hidden bar because they may not need the whole 256 busses, which also enables us to support more domains within a limited range (768MB)
Signed-off-by: Zihan Yang whois.zihan.yang@gmail.com --- src/fw/dev-q35.h | 7 +++++++ src/fw/pciinit.c | 32 ++++++++++++++++++++++++++++++++ src/hw/pci_ids.h | 1 + 3 files changed, 40 insertions(+)
diff --git a/src/fw/dev-q35.h b/src/fw/dev-q35.h index 201825d..229cd81 100644 --- a/src/fw/dev-q35.h +++ b/src/fw/dev-q35.h @@ -49,4 +49,11 @@ #define ICH9_APM_ACPI_ENABLE 0x2 #define ICH9_APM_ACPI_DISABLE 0x3
+#define PXB_PCIE_HOST_BRIDGE_MCFG_BAR 0x50 /* 64bit register */ +#define PXB_PCIE_HOST_BRIDGE_MCFG_SIZE 0x58 /* 32bit register */ +#define PXB_PCIE_HOST_BRIDGE_ENABLE Q35_HOST_BRIDGE_PCIEXBAREN +/* pxb-pcie can use [0x80000000, 0xb0000000), be careful not to overflow */ +#define PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR 0x80000000 +#define PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR_UPPER Q35_HOST_BRIDGE_PCIEXBAR_ADDR + #endif // dev-q35.h diff --git a/src/fw/pciinit.c b/src/fw/pciinit.c index c0634bc..e0ac22c 100644 --- a/src/fw/pciinit.c +++ b/src/fw/pciinit.c @@ -51,6 +51,7 @@ u64 pcimem_end = BUILD_PCIMEM_END; u64 pcimem64_start = BUILD_PCIMEM64_START; u64 pcimem64_end = BUILD_PCIMEM64_END; u64 pci_io_low_end = 0xa000; +u64 pxb_pcie_mcfg_base = PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR;
struct pci_region_entry { struct pci_device *dev; @@ -507,11 +508,42 @@ static void mch_mem_addr_setup(struct pci_device *dev, void *arg) pci_io_low_end = acpi_pm_base; }
+static void pxb_pcie_mem_addr_setup(struct pci_device *dev, void *arg) +{ + u64 mcfg_base; + u32 mcfg_size = pci_config_readl(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_SIZE); + + /* 0 means this pxb-pcie still resides in pci domain 0 */ + if (mcfg_size == 0) + return; + + if (pxb_pcie_mcfg_base + mcfg_size > + PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR_UPPER) { + dprintf(1, "PCI: Not enough space to hold new pci domains\n"); + return; + } + + mcfg_base = pxb_pcie_mcfg_base; + pxb_pcie_mcfg_base += mcfg_size; + + /* First clear old mmio, taken care of by QEMU */ + pci_config_writel(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_BAR, 0); + /* Update MCFG base */ + pci_config_writel(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_BAR + 4, + mcfg_base >> 32); + pci_config_writel(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_BAR, + (mcfg_base & 0xffffffff) | PXB_PCIE_HOST_BRIDGE_ENABLE); + + e820_add(mcfg_base, mcfg_size, E820_RESERVED); +} + static const struct pci_device_id pci_platform_tbl[] = { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82441, i440fx_mem_addr_setup), PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_Q35_MCH, mch_mem_addr_setup), + PCI_DEVICE(PCI_VENDOR_ID_REDHAT, PCI_DEVICE_ID_REDHAT_PXB_HOST, + pxb_pcie_mem_addr_setup), PCI_DEVICE_END };
diff --git a/src/hw/pci_ids.h b/src/hw/pci_ids.h index 1096461..b495920 100644 --- a/src/hw/pci_ids.h +++ b/src/hw/pci_ids.h @@ -2266,6 +2266,7 @@ #define PCI_VENDOR_ID_REDHAT 0x1b36 #define PCI_DEVICE_ID_REDHAT_ROOT_PORT 0x000C #define PCI_DEVICE_ID_REDHAT_BRIDGE 0x0001 +#define PCI_DEVICE_ID_REDHAT_PXB_HOST 0x000B
#define PCI_VENDOR_ID_TEKRAM 0x1de1 #define PCI_DEVICE_ID_TEKRAM_DC290 0xdc29
On Mon, Sep 17, 2018 at 11:02:59PM +0800, Zihan Yang wrote:
To support multiple pci domains of pxb-pcie device in qemu, we need to setup mcfg range in seabios. We use [0x80000000, 0xb0000000) to hold new domain mcfg table for now, and we need to retrieve the desired mcfg size of each pxb-pcie from a hidden bar because they may not need the whole 256 busses, which also enables us to support more domains within a limited range (768MB)
At a highlevel, this looks okay to me. I'd like to see additional reviews from others more familiar with the QEMU PCI code, though.
Is the plan to do the same thing for OVMF?
-Kevin
Signed-off-by: Zihan Yang whois.zihan.yang@gmail.com
src/fw/dev-q35.h | 7 +++++++ src/fw/pciinit.c | 32 ++++++++++++++++++++++++++++++++ src/hw/pci_ids.h | 1 + 3 files changed, 40 insertions(+)
diff --git a/src/fw/dev-q35.h b/src/fw/dev-q35.h index 201825d..229cd81 100644 --- a/src/fw/dev-q35.h +++ b/src/fw/dev-q35.h @@ -49,4 +49,11 @@ #define ICH9_APM_ACPI_ENABLE 0x2 #define ICH9_APM_ACPI_DISABLE 0x3
+#define PXB_PCIE_HOST_BRIDGE_MCFG_BAR 0x50 /* 64bit register */ +#define PXB_PCIE_HOST_BRIDGE_MCFG_SIZE 0x58 /* 32bit register */ +#define PXB_PCIE_HOST_BRIDGE_ENABLE Q35_HOST_BRIDGE_PCIEXBAREN +/* pxb-pcie can use [0x80000000, 0xb0000000), be careful not to overflow */ +#define PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR 0x80000000 +#define PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR_UPPER Q35_HOST_BRIDGE_PCIEXBAR_ADDR
#endif // dev-q35.h diff --git a/src/fw/pciinit.c b/src/fw/pciinit.c index c0634bc..e0ac22c 100644 --- a/src/fw/pciinit.c +++ b/src/fw/pciinit.c @@ -51,6 +51,7 @@ u64 pcimem_end = BUILD_PCIMEM_END; u64 pcimem64_start = BUILD_PCIMEM64_START; u64 pcimem64_end = BUILD_PCIMEM64_END; u64 pci_io_low_end = 0xa000; +u64 pxb_pcie_mcfg_base = PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR;
struct pci_region_entry { struct pci_device *dev; @@ -507,11 +508,42 @@ static void mch_mem_addr_setup(struct pci_device *dev, void *arg) pci_io_low_end = acpi_pm_base; }
+static void pxb_pcie_mem_addr_setup(struct pci_device *dev, void *arg) +{
- u64 mcfg_base;
- u32 mcfg_size = pci_config_readl(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_SIZE);
- /* 0 means this pxb-pcie still resides in pci domain 0 */
- if (mcfg_size == 0)
return;- if (pxb_pcie_mcfg_base + mcfg_size >
PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR_UPPER) {dprintf(1, "PCI: Not enough space to hold new pci domains\n");return;- }
- mcfg_base = pxb_pcie_mcfg_base;
- pxb_pcie_mcfg_base += mcfg_size;
- /* First clear old mmio, taken care of by QEMU */
- pci_config_writel(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_BAR, 0);
- /* Update MCFG base */
- pci_config_writel(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_BAR + 4,
mcfg_base >> 32);- pci_config_writel(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_BAR,
(mcfg_base & 0xffffffff) | PXB_PCIE_HOST_BRIDGE_ENABLE);- e820_add(mcfg_base, mcfg_size, E820_RESERVED);
+}
static const struct pci_device_id pci_platform_tbl[] = { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82441, i440fx_mem_addr_setup), PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_Q35_MCH, mch_mem_addr_setup),
- PCI_DEVICE(PCI_VENDOR_ID_REDHAT, PCI_DEVICE_ID_REDHAT_PXB_HOST,
pxb_pcie_mem_addr_setup),};
diff --git a/src/hw/pci_ids.h b/src/hw/pci_ids.h index 1096461..b495920 100644 --- a/src/hw/pci_ids.h +++ b/src/hw/pci_ids.h @@ -2266,6 +2266,7 @@ #define PCI_VENDOR_ID_REDHAT 0x1b36 #define PCI_DEVICE_ID_REDHAT_ROOT_PORT 0x000C #define PCI_DEVICE_ID_REDHAT_BRIDGE 0x0001 +#define PCI_DEVICE_ID_REDHAT_PXB_HOST 0x000B
#define PCI_VENDOR_ID_TEKRAM 0x1de1
#define PCI_DEVICE_ID_TEKRAM_DC290 0xdc29
2.7.4
SeaBIOS mailing list SeaBIOS@seabios.org https://mail.coreboot.org/mailman/listinfo/seabios
On 09/25/18 17:38, Kevin O'Connor wrote:
On Mon, Sep 17, 2018 at 11:02:59PM +0800, Zihan Yang wrote:
To support multiple pci domains of pxb-pcie device in qemu, we need to setup mcfg range in seabios. We use [0x80000000, 0xb0000000) to hold new domain mcfg table for now, and we need to retrieve the desired mcfg size of each pxb-pcie from a hidden bar because they may not need the whole 256 busses, which also enables us to support more domains within a limited range (768MB)
At a highlevel, this looks okay to me. I'd like to see additional reviews from others more familiar with the QEMU PCI code, though.
Is the plan to do the same thing for OVMF?
I remain entirely unconvinced that this feature is useful. (I've stated so before.)
I believe the latest QEMU RFC posting (v5) is here:
[Qemu-devel] [RFC v5 0/6] pci_expander_brdige: support separate pci domain for pxb-pcie
http://mid.mail-archive.com/1537196258-12581-1-git-send-email-whois.zihan.ya...
First, I fail to see the use case where ~256 PCI bus numbers aren't enough. If I strain myself, perhaps I can imagine using ~200 PCIe root ports on Q35 (each of which requires a separate bus number), so that we can independently hot-plug 200 devices then. And that's supposedly not enough, because we want... 300? 400? A thousand? Doesn't sound realistic to me. (This is not meant to be a strawman argument, I really have no idea what the feature would be useful for.)
Second, the v5 RFC doesn't actually address the alleged bus number shortage. IIUC, it supports a low number of ECAM ranges under 4GB, but those are (individually) limited in the bus number ranges they can accommodate (due to 32-bit address space shortage). So more or less the current approach just fragments the bus number space we already have, to multiple domains.
Third, should a subsequent iteration of the QEMU series put those extra ECAMs above 4GB, with the intent to leave the enumeration of those hierarchies to the "guest OS", it would present an incredible implementation mess for OVMF. If people gained the ability to attach storage or network to those domains, on the QEMU command line, they would expect to boot off of them, using UEFI. Then OVMF would have to make sure the controllers could be bound by their respective UEFI drivers. That in turn would require functional config space access (ECAM) at semi-random 64-bit addresses.
The layout of the 64-bit address space is already pretty darn complicated in OVMF; it depends on guest RAM size, DIMM hotplug area size, and it already dictates the base of the 64-bit MMIO aperture for BAR allocations. It also dictates how high up the DXE Core should build the page tables (you can only access 64-bit addresses if the 1:1 page tables built by the DXE Core cover them).
Obviously, UEFI on physical machines does support multiple PCI domains. There are a number of differences however:
- Both the ECAM ranges, and the MMIO apertures (for BAR allocation) of the disparate host bridges are distinct. On QEMU / OVMF, the latter part is not true (about the MMIO apertures), and this has already required us to write some nasty quirks for supposedly platform-independent core code in edk2.
- The same address ranges mentioned in the previous bullet are known in advance (they are "static"). That's a *huge* simplification opportunity to physical platform code (which is most of the time not even open source, let alone upstreamed to edk2), because the engineers can lay out the 64-bit address range, and deduce all the related artifacts from that layout, on paper, at the office.
It's a proper mess with a lot of opportunity for regressions, and I just don't see the bang for the buck.
(I didn't mean to re-hash my opinion yet again -- in the QEMU RFC v5 thread, I saw references to OVMF, stating that OVMF would not support this. I was 100% fine with those mentions, but here you asked explicitly... Some ideas just make me rant, my apologies.)
Thanks Laszlo
-Kevin
Signed-off-by: Zihan Yang whois.zihan.yang@gmail.com
src/fw/dev-q35.h | 7 +++++++ src/fw/pciinit.c | 32 ++++++++++++++++++++++++++++++++ src/hw/pci_ids.h | 1 + 3 files changed, 40 insertions(+)
diff --git a/src/fw/dev-q35.h b/src/fw/dev-q35.h index 201825d..229cd81 100644 --- a/src/fw/dev-q35.h +++ b/src/fw/dev-q35.h @@ -49,4 +49,11 @@ #define ICH9_APM_ACPI_ENABLE 0x2 #define ICH9_APM_ACPI_DISABLE 0x3
+#define PXB_PCIE_HOST_BRIDGE_MCFG_BAR 0x50 /* 64bit register */ +#define PXB_PCIE_HOST_BRIDGE_MCFG_SIZE 0x58 /* 32bit register */ +#define PXB_PCIE_HOST_BRIDGE_ENABLE Q35_HOST_BRIDGE_PCIEXBAREN +/* pxb-pcie can use [0x80000000, 0xb0000000), be careful not to overflow */ +#define PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR 0x80000000 +#define PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR_UPPER Q35_HOST_BRIDGE_PCIEXBAR_ADDR
#endif // dev-q35.h diff --git a/src/fw/pciinit.c b/src/fw/pciinit.c index c0634bc..e0ac22c 100644 --- a/src/fw/pciinit.c +++ b/src/fw/pciinit.c @@ -51,6 +51,7 @@ u64 pcimem_end = BUILD_PCIMEM_END; u64 pcimem64_start = BUILD_PCIMEM64_START; u64 pcimem64_end = BUILD_PCIMEM64_END; u64 pci_io_low_end = 0xa000; +u64 pxb_pcie_mcfg_base = PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR;
struct pci_region_entry { struct pci_device *dev; @@ -507,11 +508,42 @@ static void mch_mem_addr_setup(struct pci_device *dev, void *arg) pci_io_low_end = acpi_pm_base; }
+static void pxb_pcie_mem_addr_setup(struct pci_device *dev, void *arg) +{
- u64 mcfg_base;
- u32 mcfg_size = pci_config_readl(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_SIZE);
- /* 0 means this pxb-pcie still resides in pci domain 0 */
- if (mcfg_size == 0)
return;- if (pxb_pcie_mcfg_base + mcfg_size >
PXB_PCIE_HOST_BRIDGE_MCFG_SIZE_ADDR_UPPER) {dprintf(1, "PCI: Not enough space to hold new pci domains\n");return;- }
- mcfg_base = pxb_pcie_mcfg_base;
- pxb_pcie_mcfg_base += mcfg_size;
- /* First clear old mmio, taken care of by QEMU */
- pci_config_writel(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_BAR, 0);
- /* Update MCFG base */
- pci_config_writel(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_BAR + 4,
mcfg_base >> 32);- pci_config_writel(dev->bdf, PXB_PCIE_HOST_BRIDGE_MCFG_BAR,
(mcfg_base & 0xffffffff) | PXB_PCIE_HOST_BRIDGE_ENABLE);- e820_add(mcfg_base, mcfg_size, E820_RESERVED);
+}
static const struct pci_device_id pci_platform_tbl[] = { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82441, i440fx_mem_addr_setup), PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_Q35_MCH, mch_mem_addr_setup),
- PCI_DEVICE(PCI_VENDOR_ID_REDHAT, PCI_DEVICE_ID_REDHAT_PXB_HOST,
pxb_pcie_mem_addr_setup),};
diff --git a/src/hw/pci_ids.h b/src/hw/pci_ids.h index 1096461..b495920 100644 --- a/src/hw/pci_ids.h +++ b/src/hw/pci_ids.h @@ -2266,6 +2266,7 @@ #define PCI_VENDOR_ID_REDHAT 0x1b36 #define PCI_DEVICE_ID_REDHAT_ROOT_PORT 0x000C #define PCI_DEVICE_ID_REDHAT_BRIDGE 0x0001 +#define PCI_DEVICE_ID_REDHAT_PXB_HOST 0x000B
#define PCI_VENDOR_ID_TEKRAM 0x1de1
#define PCI_DEVICE_ID_TEKRAM_DC290 0xdc29
2.7.4
SeaBIOS mailing list SeaBIOS@seabios.org https://mail.coreboot.org/mailman/listinfo/seabios
Hi,
Second, the v5 RFC doesn't actually address the alleged bus number shortage. IIUC, it supports a low number of ECAM ranges under 4GB, but those are (individually) limited in the bus number ranges they can accommodate (due to 32-bit address space shortage). So more or less the current approach just fragments the bus number space we already have, to multiple domains.
Havn't looked at the qemu side too close yet, but as I understand things the firmware programs the ECAM location (simliar to the q35 mmconf bar), and this is just a limitation of the current seabios patch.
So, no, *that* part wouldn't be messy in ovmf, you can simply place the ECAMs where you want.
cheers, Gerd
On 09/26/18 06:44, Gerd Hoffmann wrote:
Hi,
Second, the v5 RFC doesn't actually address the alleged bus number shortage. IIUC, it supports a low number of ECAM ranges under 4GB, but those are (individually) limited in the bus number ranges they can accommodate (due to 32-bit address space shortage). So more or less the current approach just fragments the bus number space we already have, to multiple domains.
Havn't looked at the qemu side too close yet, but as I understand things the firmware programs the ECAM location (simliar to the q35 mmconf bar), and this is just a limitation of the current seabios patch.
So, no, *that* part wouldn't be messy in ovmf, you can simply place the ECAMs where you want.
Figuring out "wherever I want" is the problem. It's not simple. The 64-bit MMIO aperture (for BAR allocation) can also be placed mostly "wherever the firmware wants", and that wasn't simple either. All these things end up depending on each other.
https://bugzilla.redhat.com/show_bug.cgi?id=1353591#c8
(
The placement of the q35 MMCONF BAR was difficult too; I needed your help with the low RAM split that QEMU would choose.
v1 discussion:
http://mid.mail-archive.com/1457340448.25423.43.camel@redhat.com
v2 patch (ended up as commit 7b8fe63561b4):
http://mid.mail-archive.com/1457446804-18892-4-git-send-email-lersek@redhat....
These things add up :(
)
Laszlo
On Wed, Sep 26, 2018 at 10:47:42AM +0200, Laszlo Ersek wrote:
On 09/26/18 06:44, Gerd Hoffmann wrote:
Hi,
Second, the v5 RFC doesn't actually address the alleged bus number shortage. IIUC, it supports a low number of ECAM ranges under 4GB, but those are (individually) limited in the bus number ranges they can accommodate (due to 32-bit address space shortage). So more or less the current approach just fragments the bus number space we already have, to multiple domains.
Havn't looked at the qemu side too close yet, but as I understand things the firmware programs the ECAM location (simliar to the q35 mmconf bar), and this is just a limitation of the current seabios patch.
So, no, *that* part wouldn't be messy in ovmf, you can simply place the ECAMs where you want.
Figuring out "wherever I want" is the problem. It's not simple. The 64-bit MMIO aperture (for BAR allocation) can also be placed mostly "wherever the firmware wants", and that wasn't simple either. All these things end up depending on each other.
Agree. Due to the very dynamic nature virtual hardware in qemu ovmf has to adapt to *alot* more stuff at runtime than physical hardware platforms, where you for the most part know what hardware is there and can handle alot of things at compile time (without nasty initialization order issues).
Physical address space bits not being reliable is rather bad too. It makes address space management harder, because we squeeze everything below 64G if possible. Just in case, we might have 36 physbits only.
It's the reason we need need stuff like address space reservation for memory hotplug in the first place, instead of simply reserving 1/4 or 1/8 of the physical address space at the topmost location for the 64bit mmio aperture.
Oh, and after the rant back to the original topic: I just wanted point out that you don't have to be worryed about all the 32bit address space issues described in the v5 cover letter. That is how the seabios patch handles things. OVMF can handle things differently. For example totally ignore pxb-pcie in 32bit builds, and place the ECAMs above 4G (take away some room from the 64-bit MMIO aperture) unconditionally in 64bit builds.
cheers, Gerd
HI Laszlo Laszlo Ersek lersek@redhat.com 于2018年9月26日周三 上午1:17写道：
On 09/25/18 17:38, Kevin O'Connor wrote:
On Mon, Sep 17, 2018 at 11:02:59PM +0800, Zihan Yang wrote:
To support multiple pci domains of pxb-pcie device in qemu, we need to setup mcfg range in seabios. We use [0x80000000, 0xb0000000) to hold new domain mcfg table for now, and we need to retrieve the desired mcfg size of each pxb-pcie from a hidden bar because they may not need the whole 256 busses, which also enables us to support more domains within a limited range (768MB)
At a highlevel, this looks okay to me. I'd like to see additional reviews from others more familiar with the QEMU PCI code, though.
Is the plan to do the same thing for OVMF?
I remain entirely unconvinced that this feature is useful. (I've stated so before.)
I believe the latest QEMU RFC posting (v5) is here:
[Qemu-devel] [RFC v5 0/6] pci_expander_brdige: support separate pci domain for pxb-pcie
http://mid.mail-archive.com/1537196258-12581-1-git-send-email-whois.zihan.ya...
First, I fail to see the use case where ~256 PCI bus numbers aren't enough. If I strain myself, perhaps I can imagine using ~200 PCIe root ports on Q35 (each of which requires a separate bus number), so that we can independently hot-plug 200 devices then. And that's supposedly not enough, because we want... 300? 400? A thousand? Doesn't sound realistic to me. (This is not meant to be a strawman argument, I really have no idea what the feature would be useful for.)
It might not be very intuitive, but it indeed exists. The very beginning discussion about 4 months ago has mentioned a possible use case, and I paste it here
- We have Ray from Intel trying to use 1000 virtio-net devices - We may have a VM managing some backups (tapes), we may have a lot of these. - We may want indeed to create a nested solution as Michael mentioned.
The thread can be found in https://lists.gnu.org/archive/html/qemu-devel/2018-05/msg04667.html
Also, a later post from a person in Dell stated in the list that he would need this feature for Intel VMD In Dell EMC? I have no idea about the details, but since they went here for help, I guess they do can benefit from it somehow.
Second, the v5 RFC doesn't actually address the alleged bus number shortage. IIUC, it supports a low number of ECAM ranges under 4GB, but those are (individually) limited in the bus number ranges they can accommodate (due to 32-bit address space shortage). So more or less the current approach just fragments the bus number space we already have, to multiple domains.
Third, should a subsequent iteration of the QEMU series put those extra ECAMs above 4GB, with the intent to leave the enumeration of those hierarchies to the "guest OS", it would present an incredible implementation mess for OVMF. If people gained the ability to attach storage or network to those domains, on the QEMU command line, they would expect to boot off of them, using UEFI. Then OVMF would have to make sure the controllers could be bound by their respective UEFI drivers. That in turn would require functional config space access (ECAM) at semi-random 64-bit addresses.
I'm not familiar with OVMF, so I'm afraid I don't know how to make it easier for OVMF, the division of 64bit space in OVMF is out of my purpose. There is no plan to implement it in OVMF for now, we just want to make the seabios/qemu patch a proof of concept.
As for the seabios, it access devices through port 0xcf8/0xcfc, which is binded to q35 host in qemu. If we want to change the mmconfig size of pxb-pcie (instead of using whole 256MB), we must know its desired size, which is passed as a hidden bar. Unfortunately the configuration space of pxb-pcie device cannot be accessed with 0xcf8/0xcfc because they are in different host bridge. At this time the ECAM is not configured yet, so no able to use MMIO too. In previous version, I tried to bind pxb host to other ports in qemu, so that we can use port io to access the config space of pxb-pcie, but it seems a little dirty,
Another issue is how seabios initialize things. It will first do pci_setup when things like RSDP is not loaded. It is inconvenient to retrieve MCFG table and other information, so we cannot infer the mmconfig addr and size from MCFG table in seabios.
Therefore we fall back to an alternative that we support 4x of devices as the first step, and let the guest os do the initialization. The inability to boot from devices in another domain is indeed an issue, and we don't have very good solution to it yet.
Things might change in the future if we can figure out a better solution, and I hope we can have an easier and more elegant solution in OVMF. But now we are just trying to give a possible solution as a poc.
Thanks Zihan
-
Gerd Hoffmann -
Kevin O'Connor -
Laszlo Ersek -
Zihan Yang