[U-Boot] [PATCH 1/2] dm/pci: Change the first CFG read to Vendor ID in enumeration
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com --- drivers/pci/pci-uclass.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/drivers/pci/pci-uclass.c b/drivers/pci/pci-uclass.c index eb118f3496..c8cb9cdad9 100644 --- a/drivers/pci/pci-uclass.c +++ b/drivers/pci/pci-uclass.c @@ -748,16 +748,19 @@ int pci_bind_bus_devices(struct udevice *bus) found_multi = false; if (PCI_FUNC(bdf) && !found_multi) continue; + /* Check only the first access, we don't expect problems */ - ret = pci_bus_read_config(bus, bdf, PCI_HEADER_TYPE, - &header_type, PCI_SIZE_8); + ret = pci_bus_read_config(bus, bdf, PCI_VENDOR_ID, &vendor, + PCI_SIZE_16); if (ret) goto error; - pci_bus_read_config(bus, bdf, PCI_VENDOR_ID, &vendor, - PCI_SIZE_16); + if (vendor == 0xffff || vendor == 0x0000) continue;
+ pci_bus_read_config(bus, bdf, PCI_HEADER_TYPE, + &header_type, PCI_SIZE_8); + if (!PCI_FUNC(bdf)) found_multi = header_type & 0x80;
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com --- drivers/pci/pci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index 221054593e..590f7b123a 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -361,13 +361,13 @@ int pci_hose_scan_bus(struct pci_controller *hose, int bus) if (PCI_FUNC(dev) && !found_multi) continue;
- pci_hose_read_config_byte(hose, dev, PCI_HEADER_TYPE, &header_type); - pci_hose_read_config_word(hose, dev, PCI_VENDOR_ID, &vendor);
if (vendor == 0xffff || vendor == 0x0000) continue;
+ pci_hose_read_config_byte(hose, dev, PCI_HEADER_TYPE, &header_type); + if (!PCI_FUNC(dev)) found_multi = header_type & 0x80;
On Mon, Oct 8, 2018 at 4:36 PM Zhiqiang Hou Zhiqiang.Hou@nxp.com wrote:
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com
drivers/pci/pci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Simon, do we need fix the non-DM version or adding new "capability" to the non-DM drivers? I guess the answer is no and we should set some deadline for PCI DM conversion too?
Regards, Bin
Hi Bin,
On 8 October 2018 at 02:54, Bin Meng bmeng.cn@gmail.com wrote:
On Mon, Oct 8, 2018 at 4:36 PM Zhiqiang Hou Zhiqiang.Hou@nxp.com wrote:
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com
drivers/pci/pci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Simon, do we need fix the non-DM version or adding new "capability" to the non-DM drivers? I guess the answer is no and we should set some deadline for PCI DM conversion too?
We should not fix or add new features to non-DM PCI. No one should be using it at this point.
I agree about the deadline. Do you want to send a patch?
Regards, Simon
Hi Simon,
On Tue, Oct 9, 2018 at 11:41 AM Simon Glass sjg@chromium.org wrote:
Hi Bin,
On 8 October 2018 at 02:54, Bin Meng bmeng.cn@gmail.com wrote:
On Mon, Oct 8, 2018 at 4:36 PM Zhiqiang Hou Zhiqiang.Hou@nxp.com wrote:
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com
drivers/pci/pci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Simon, do we need fix the non-DM version or adding new "capability" to the non-DM drivers? I guess the answer is no and we should set some deadline for PCI DM conversion too?
We should not fix or add new features to non-DM PCI. No one should be using it at this point.
I agree about the deadline. Do you want to send a patch?
Actually I am not sure how to send a patch for such kind of deadline stuff :-) Please go ahead.
Regards, Bin
Hi Simon and Bin,
Do I need to send the standalone patch 1/2 in version 2?
Thanks, Zhiqiang
-----Original Message----- From: sjg@google.com sjg@google.com On Behalf Of Simon Glass Sent: 2018年10月9日 11:41 To: Bin Meng bmeng.cn@gmail.com Cc: Z.q. Hou zhiqiang.hou@nxp.com; U-Boot Mailing List u-boot@lists.denx.de; Christian Gmeiner christian.gmeiner@gmail.com; Tuomas Tynkkynen tuomas.tynkkynen@iki.fi; Bernhard Messerklinger bernhard.messerklinger@br-automation.com; Masahiro Yamada yamada.masahiro@socionext.com; Mingkai Hu mingkai.hu@nxp.com; M.h. Lian minghuan.lian@nxp.com Subject: Re: [PATCH 2/2] pci: Change the first CFG read to Vendor ID in enumeration
Hi Bin,
On 8 October 2018 at 02:54, Bin Meng bmeng.cn@gmail.com wrote:
On Mon, Oct 8, 2018 at 4:36 PM Zhiqiang Hou Zhiqiang.Hou@nxp.com
wrote:
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com
drivers/pci/pci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Simon, do we need fix the non-DM version or adding new "capability" to the non-DM drivers? I guess the answer is no and we should set some deadline for PCI DM conversion too?
We should not fix or add new features to non-DM PCI. No one should be using it at this point.
I agree about the deadline. Do you want to send a patch?
Regards, Simon
On 9 October 2018 at 02:07, Z.q. Hou zhiqiang.hou@nxp.com wrote:
Hi Simon and Bin,
Do I need to send the standalone patch 1/2 in version 2?
I don't think that is needed, no.
- SImon
Thanks, Zhiqiang
-----Original Message----- From: sjg@google.com sjg@google.com On Behalf Of Simon Glass Sent: 2018年10月9日 11:41 To: Bin Meng bmeng.cn@gmail.com Cc: Z.q. Hou zhiqiang.hou@nxp.com; U-Boot Mailing List u-boot@lists.denx.de; Christian Gmeiner christian.gmeiner@gmail.com; Tuomas Tynkkynen tuomas.tynkkynen@iki.fi; Bernhard Messerklinger bernhard.messerklinger@br-automation.com; Masahiro Yamada yamada.masahiro@socionext.com; Mingkai Hu mingkai.hu@nxp.com; M.h. Lian minghuan.lian@nxp.com Subject: Re: [PATCH 2/2] pci: Change the first CFG read to Vendor ID in enumeration
Hi Bin,
On 8 October 2018 at 02:54, Bin Meng bmeng.cn@gmail.com wrote:
On Mon, Oct 8, 2018 at 4:36 PM Zhiqiang Hou Zhiqiang.Hou@nxp.com
wrote:
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com
drivers/pci/pci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Simon, do we need fix the non-DM version or adding new "capability" to the non-DM drivers? I guess the answer is no and we should set some deadline for PCI DM conversion too?
We should not fix or add new features to non-DM PCI. No one should be using it at this point.
I agree about the deadline. Do you want to send a patch?
Regards, Simon
On Mon, Oct 8, 2018 at 4:36 PM Zhiqiang Hou Zhiqiang.Hou@nxp.com wrote:
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com
drivers/pci/pci-uclass.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-)
Reviewed-by: Bin Meng bmeng.cn@gmail.com
Hi Bin,
Thanks a lot for your review!
Thanks, Zhiqiang
-----Original Message----- From: Bin Meng bmeng.cn@gmail.com Sent: 2018年10月8日 16:53 To: Z.q. Hou zhiqiang.hou@nxp.com Cc: U-Boot Mailing List u-boot@lists.denx.de; Simon Glass sjg@chromium.org; Christian Gmeiner christian.gmeiner@gmail.com; Tuomas Tynkkynen tuomas.tynkkynen@iki.fi; Bernhard Messerklinger bernhard.messerklinger@br-automation.com; Masahiro Yamada yamada.masahiro@socionext.com; Mingkai Hu mingkai.hu@nxp.com; M.h. Lian minghuan.lian@nxp.com Subject: Re: [PATCH 1/2] dm/pci: Change the first CFG read to Vendor ID in enumeration
On Mon, Oct 8, 2018 at 4:36 PM Zhiqiang Hou Zhiqiang.Hou@nxp.com wrote:
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com
drivers/pci/pci-uclass.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-)
Reviewed-by: Bin Meng bmeng.cn@gmail.com
Hi Bin,
Thanks a lot for your review!
Thanks, Zhiqiang
-----Original Message----- From: Bin Meng bmeng.cn@gmail.com Sent: 2018年10月8日 16:53 To: Z.q. Hou zhiqiang.hou@nxp.com Cc: U-Boot Mailing List u-boot@lists.denx.de; Simon Glass sjg@chromium.org; Christian Gmeiner christian.gmeiner@gmail.com; Tuomas Tynkkynen tuomas.tynkkynen@iki.fi; Bernhard Messerklinger bernhard.messerklinger@br-automation.com; Masahiro Yamada yamada.masahiro@socionext.com; Mingkai Hu mingkai.hu@nxp.com; M.h. Lian minghuan.lian@nxp.com Subject: Re: [PATCH 1/2] dm/pci: Change the first CFG read to Vendor ID in enumeration
On Mon, Oct 8, 2018 at 4:36 PM Zhiqiang Hou Zhiqiang.Hou@nxp.com wrote:
From: Hou Zhiqiang Zhiqiang.Hou@nxp.com
As the PCIe specification recommend reading the Vendor ID register to determine if a Function is present, read the Vendor ID of a non-existent Function must not result in system error, so we'd better make the first CFG read to Vendor ID instead of Header Type register in the PCIe enumeration.
Signed-off-by: Hou Zhiqiang Zhiqiang.Hou@nxp.com
drivers/pci/pci-uclass.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-)
Reviewed-by: Bin Meng bmeng.cn@gmail.com
Applied to u-boot-dm/master, thanks!
participants (5)
-
Bin Meng -
Simon Glass -
sjg＠google.com -
Z.q. Hou -
Zhiqiang Hou