Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.31 as permitted sender) client-ip=23.128.96.31;
Feedback-ID: i1568416f:Fastmail
From:   =?UTF-8?q?Marek=20Marczykowski-G=C3=B3recki?= 
        <marmarek@invisiblethingslab.com>
To:     linux-kernel@vger.kernel.org
Cc:     =?UTF-8?q?Marek=20Marczykowski-G=C3=B3recki?= 
        <marmarek@invisiblethingslab.com>, Juergen Gross <jgross@suse.com>,
        Stefano Stabellini <sstabellini@kernel.org>,
        Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>,
        Jan Beulich <jbeulich@suse.com>,
        xen-devel@lists.xenproject.org (moderated list:XEN HYPERVISOR INTERFACE)
Subject: [PATCH v4] xen-pciback: Consider INTx disabled when MSI/MSI-X is enabled
Date:   Mon, 16 Oct 2023 15:13:25 +0200
Message-ID: <20231016131348.1734721-1-marmarek@invisiblethingslab.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk

Linux enables MSI-X before disabling INTx, but keeps MSI-X masked until
the table is filled. Then it disables INTx just before clearing MASKALL
bit. Currently this approach is rejected by xen-pciback.
According to the PCIe spec, device cannot use INTx when MSI/MSI-X is
enabled (in other words: enabling MSI/MSI-X implicitly disables INTx).

Change the logic to consider INTx disabled if MSI/MSI-X is enabled. This
applies to three places:
 - checking currently enabled interrupts type,
 - transition to MSI/MSI-X - where INTx would be implicitly disabled,
 - clearing INTx disable bit - which can be allowed even if MSI/MSI-X is
   enabled, as device should consider INTx disabled anyway in that case

Fixes: 5e29500eba2a ("xen-pciback: Allow setting PCI_MSIX_FLAGS_MASKALL too")
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Acked-by: Juergen Gross <jgross@suse.com>
---
Changes in v4:
 - update comment (retain A-b)
Changes in v3:
 - allow clearing INTx regardless of MSI/MSI-X state, to be consistent
   with enabling MSI/MSI-X
Changes in v2:
 - restructure the patch to consider not only MASKALL bit, but enabling
   MSI/MSI-X generally, without explicitly disabling INTx first
---
 drivers/xen/xen-pciback/conf_space.c          | 19 +++++++++++------
 .../xen/xen-pciback/conf_space_capability.c   |  8 ++++++-
 drivers/xen/xen-pciback/conf_space_header.c   | 21 +++----------------
 3 files changed, 23 insertions(+), 25 deletions(-)

diff --git a/drivers/xen/xen-pciback/conf_space.c b/drivers/xen/xen-pciback/conf_space.c
index 059de92aea7d..d47eee6c5143 100644
--- a/drivers/xen/xen-pciback/conf_space.c
+++ b/drivers/xen/xen-pciback/conf_space.c
@@ -288,12 +288,6 @@ int xen_pcibk_get_interrupt_type(struct pci_dev *dev)
 	u16 val;
 	int ret = 0;
 
-	err = pci_read_config_word(dev, PCI_COMMAND, &val);
-	if (err)
-		return err;
-	if (!(val & PCI_COMMAND_INTX_DISABLE))
-		ret |= INTERRUPT_TYPE_INTX;
-
 	/*
 	 * Do not trust dev->msi(x)_enabled here, as enabling could be done
 	 * bypassing the pci_*msi* functions, by the qemu.
@@ -316,6 +310,19 @@ int xen_pcibk_get_interrupt_type(struct pci_dev *dev)
 		if (val & PCI_MSIX_FLAGS_ENABLE)
 			ret |= INTERRUPT_TYPE_MSIX;
 	}
+
+	/*
+	 * PCIe spec says device cannot use INTx if MSI/MSI-X is enabled,
+	 * so check for INTx only when both are disabled.
+	 */
+	if (!ret) {
+		err = pci_read_config_word(dev, PCI_COMMAND, &val);
+		if (err)
+			return err;
+		if (!(val & PCI_COMMAND_INTX_DISABLE))
+			ret |= INTERRUPT_TYPE_INTX;
+	}
+
 	return ret ?: INTERRUPT_TYPE_NONE;
 }
 
diff --git a/drivers/xen/xen-pciback/conf_space_capability.c b/drivers/xen/xen-pciback/conf_space_capability.c
index 097316a74126..1948a9700c8f 100644
--- a/drivers/xen/xen-pciback/conf_space_capability.c
+++ b/drivers/xen/xen-pciback/conf_space_capability.c
@@ -236,10 +236,16 @@ static int msi_msix_flags_write(struct pci_dev *dev, int offset, u16 new_value,
 		return PCIBIOS_SET_FAILED;
 
 	if (new_value & field_config->enable_bit) {
-		/* don't allow enabling together with other interrupt types */
+		/*
+		 * Don't allow enabling together with other interrupt type, but do
+		 * allow enabling MSI(-X) while INTx is still active to please Linuxes
+		 * MSI(-X) startup sequence. It is safe to do, as according to PCI
+		 * spec, device with enabled MSI(-X) shouldn't use INTx.
+		 */
 		int int_type = xen_pcibk_get_interrupt_type(dev);
 
 		if (int_type == INTERRUPT_TYPE_NONE ||
+		    int_type == INTERRUPT_TYPE_INTX ||
 		    int_type == field_config->int_type)
 			goto write;
 		return PCIBIOS_SET_FAILED;
diff --git a/drivers/xen/xen-pciback/conf_space_header.c b/drivers/xen/xen-pciback/conf_space_header.c
index 981435103af1..fc0332645966 100644
--- a/drivers/xen/xen-pciback/conf_space_header.c
+++ b/drivers/xen/xen-pciback/conf_space_header.c
@@ -104,24 +104,9 @@ static int command_write(struct pci_dev *dev, int offset, u16 value, void *data)
 		pci_clear_mwi(dev);
 	}
 
-	if (dev_data && dev_data->allow_interrupt_control) {
-		if ((cmd->val ^ value) & PCI_COMMAND_INTX_DISABLE) {
-			if (value & PCI_COMMAND_INTX_DISABLE) {
-				pci_intx(dev, 0);
-			} else {
-				/* Do not allow enabling INTx together with MSI or MSI-X. */
-				switch (xen_pcibk_get_interrupt_type(dev)) {
-				case INTERRUPT_TYPE_NONE:
-					pci_intx(dev, 1);
-					break;
-				case INTERRUPT_TYPE_INTX:
-					break;
-				default:
-					return PCIBIOS_SET_FAILED;
-				}
-			}
-		}
-	}
+	if (dev_data && dev_data->allow_interrupt_control &&
+	    ((cmd->val ^ value) & PCI_COMMAND_INTX_DISABLE))
+		pci_intx(dev, !(value & PCI_COMMAND_INTX_DISABLE));
 
 	cmd->val = value;
 
-- 
2.41.0