kernel/git/stable/linux.git/drivers/pci, branch linux-7.1.y

PCI: host-generic: Fix NULL pointer dereference on 32-bit CAM systems

2026-08-27T12:34:34+00:00

commit 008cb88edb41f3c7c8e0ed763ff9f26719830984 upstream. On 32-bit systems the config space is too large to ioremap in one go, so pci_ecam_create() maps each bus segment separately and relies on the ->add_bus callback (pci_ecam_add_bus) to populate the per-bus mapping in cfg->winp[]. pci_ecam_map_bus() then uses that mapping as the base for every config access. The generic ECAM ops (pci_generic_ecam_ops) already provide the ->add_bus and ->remove_bus callbacks, but the CAM (legacy) ops in pci-host-generic.c do not. As a result, on a 32-bit host using "pci-host-cam-generic" the per-bus mapping is never set up and the first config read dereferences a NULL base, crashing during bus enumeration: Unable to handle kernel NULL pointer dereference at virtual address 00000800 Oops [#1] CPU: 0 PID: 1 Comm: swapper Not tainted 6.9.7+ #43 Hardware name: Digilent Nexys-Video-A7 RV32 (DT) epc : pci_generic_config_read+0x40/0xb0 ra : pci_generic_config_read+0x2c/0xb0 [] pci_generic_config_read+0x40/0xb0 [] pci_bus_read_config_dword+0x50/0xb0 [] pci_bus_generic_read_dev_vendor_id+0x3c/0x1ec [] pci_scan_single_device+0xa4/0x11c [] pci_scan_slot+0x9c/0x23c [] pci_scan_child_bus_extend+0x58/0x2f4 [] pci_scan_root_bus_bridge+0x64/0xe8 [] pci_host_probe+0x20/0xc8 [] pci_host_common_probe+0x144/0x1e4 Fix this by giving the CAM ops the same ->add_bus/->remove_bus callbacks. Since pci_ecam_add_bus() and pci_ecam_remove_bus() are static to ecam.c, move the CAM ops definition there as pci_generic_cam_ops (mirroring pci_generic_ecam_ops) and export it for pci-host-generic.c to reference. Fixes: 8fe55ef23387 ("PCI: Dynamically map ECAM regions") Signed-off-by: Steffen Persvold [mani: removed timestamp from log] Signed-off-by: Manivannan Sadhasivam Cc: stable@vger.kernel.org Link: https://patch.msgid.link/20260709122446.3151899-1-spersvold@gmail.com Signed-off-by: Greg Kroah-Hartman

PCI: imx6: Keep i.MX6 Root Port MSI/MSI-X Capabilities with iMSI-RX to work around hardware bug

2026-08-09T18:26:45+00:00

commit b4bee12ebeccfcd5959ace2c3a4af08f5a917d4c upstream. On some NXP chipsets, disabling Root Port MSI/MSI-X Capabilities blocks MSIs originating from Endpoints from reaching the iMSI-RX controller. To address this hardware bug, commit 3a4e8302e72f ("PCI: imx6: Keep Root Port MSI capability with iMSI-RX to work around hardware bug") preserves Root Port MSI and MSI-X Capabilities on i.MX7D, i.MX8MM, and i.MX8MQ when iMSI-RX is in use. The same applies to i.MX6Q, i.MX6QP, and i.MX6SX, so preserve Root Port MSI/MSI-X Capabilities there as well. Note that preserving these Capabilities means Root Port-originated MSIs such as AER and PME won't be received due to separate hardware limitations. Users may need to use workarounds such as passing the 'pcie_pme=nomsi' command-line parameter. Fixes: f5cd8a929c825 ("PCI: dwc: Remove MSI/MSIX capability for Root Port if iMSI-RX is used as MSI controller") Signed-off-by: Soeren Moch Signed-off-by: Richard Zhu Signed-off-by: Bjorn Helgaas Reviewed-by: Frank Li Reviewed-by: Manivannan Sadhasivam Acked-by: Richard Zhu Cc: stable@vger.kernel.org # 7.0+ Link: https://patch.msgid.link/20260717033203.2965045-1-hongxing.zhu@oss.nxp.com Signed-off-by: Greg Kroah-Hartman

PCI: endpoint: pci-epf-vntb: Exclude reserved slots from db_valid_mask

2026-07-24T14:20:27+00:00

[ Upstream commit 823468a4ea1613f1c1235bd16af8b9c6eb9c9677 ] In pci-epf-vntb, db_count represents the total number of doorbell slots exposed to the peer, including: - slot #0 reserved for link events, and - slot #1 historically unused (kept for compatibility). Only the remaining slots correspond to actual doorbell bits. The current db_valid_mask() exposes all slots as valid doorbells. Limit db_valid_mask() to the real doorbell bits by returning BIT_ULL(db_count - 2) - 1, and guard against db_count < 2. Fixes: e35f56bb0330 ("PCI: endpoint: Support NTB transfer between RC and EP") Signed-off-by: Koichiro Den Signed-off-by: Manivannan Sadhasivam Signed-off-by: Bjorn Helgaas Reviewed-by: Frank Li Link: https://patch.msgid.link/20260513024923.451765-7-den@valinux.co.jp Signed-off-by: Sasha Levin

PCI: endpoint: pci-epf-vntb: Report 0-based doorbell vector via ntb_db_event()

2026-07-24T14:20:24+00:00

[ Upstream commit 91fb4488cd615a39360bc4160a10cb3236189ba1 ] ntb_db_event() expects the vector number to be relative to the first doorbell vector starting at 0. pci-epf-vntb reserves vector 0 for link events and uses higher vector indices for doorbells. By passing the raw slot index to ntb_db_event(), it effectively assumes that doorbell 0 maps to vector 1. However, because the host uses a legacy slot layout and writes doorbell 0 into the third slot, doorbell 0 ultimately appears as vector 2 from the NTB core perspective. Adjust pci-epf-vntb to: - skip the unused second slot, and - report doorbells as 0-based vectors (DB#0 -> vector 0). This change does not introduce a behavioral difference until .db_vector_count()/.db_vector_mask() are implemented, because without those callbacks NTB clients effectively ignore the vector number. Fixes: e35f56bb0330 ("PCI: endpoint: Support NTB transfer between RC and EP") Signed-off-by: Koichiro Den Signed-off-by: Manivannan Sadhasivam Signed-off-by: Bjorn Helgaas Reviewed-by: Frank Li Link: https://patch.msgid.link/20260513024923.451765-4-den@valinux.co.jp Signed-off-by: Sasha Levin

PCI: endpoint: pci-epf-vntb: Defer pci_epc_raise_irq() out of atomic context

2026-07-24T14:20:24+00:00

[ Upstream commit 18355c1e986582aaff2c488b1a2ce79bac8f3cf9 ] The NTB .peer_db_set() callback may be invoked from atomic context. pci-epf-vntb currently calls pci_epc_raise_irq() directly, but pci_epc_raise_irq() may sleep (it takes epc->lock). Avoid sleeping in atomic context by coalescing doorbell bits into an atomic64 pending mask and raising MSIs from a work item. Limit the amount of work per run to avoid monopolizing the workqueue under a doorbell storm. Clear stale pending bits before enabling the work item and after disabling it during cleanup. Also mask requested doorbells against the currently valid doorbell mask before queueing work, and iterate the pending u64 with __ffs64() so high doorbell bits are handled correctly. Fixes: e35f56bb0330 ("PCI: endpoint: Support NTB transfer between RC and EP") Signed-off-by: Koichiro Den Signed-off-by: Manivannan Sadhasivam Signed-off-by: Bjorn Helgaas Reviewed-by: Frank Li Link: https://patch.msgid.link/20260513024923.451765-3-den@valinux.co.jp Stable-dep-of: 91fb4488cd61 ("PCI: endpoint: pci-epf-vntb: Report 0-based doorbell vector via ntb_db_event()") Signed-off-by: Sasha Levin

PCI: endpoint: pci-epf-vntb: Document legacy MSI doorbell offset

2026-07-24T14:20:24+00:00

[ Upstream commit 6a7db4fcc6c23b1d25e676400d764bdcb7dc1ccb ] vntb_epf_peer_db_set() raises an MSI interrupt to notify the RC side of a doorbell event. pci_epc_raise_irq(..., PCI_IRQ_MSI, interrupt_num) takes a 1-based MSI interrupt number. The ntb_hw_epf driver reserves MSI #1 for link events, so doorbells would naturally start at MSI #2 (doorbell bit 0 -> MSI #2). However, pci-epf-vntb has historically applied an extra offset and mapped doorbell bit 0 to MSI #3. This matches the legacy behavior of ntb_hw_epf and has been preserved since commit e35f56bb0330 ("PCI: endpoint: Support NTB transfer between RC and EP"). This offset has not surfaced as a functional issue because: - ntb_hw_epf typically allocates enough MSI vectors, so the off-by-one still hits a valid MSI vector, and - ntb_hw_epf does not implement .db_vector_count()/.db_vector_mask(), so client drivers such as ntb_transport effectively ignore the vector number and schedule all QPs. Correcting the MSI number would break interoperability with peers running older kernels. Document the legacy offset to avoid confusion when enabling per-db-vector handling. Signed-off-by: Koichiro Den Signed-off-by: Manivannan Sadhasivam Signed-off-by: Bjorn Helgaas Reviewed-by: Frank Li Link: https://patch.msgid.link/20260513024923.451765-2-den@valinux.co.jp Stable-dep-of: 91fb4488cd61 ("PCI: endpoint: pci-epf-vntb: Report 0-based doorbell vector via ntb_db_event()") Signed-off-by: Sasha Levin

PCI: endpoint: pci-epf-ntb: Add check to detect 'db_count' value of 0

2026-07-24T14:20:24+00:00

[ Upstream commit d1db6d7c2485ee475a04d8354fbb5b26ea10d978 ] epf_ntb->db_count value should be within 1 to MAX_DB_COUNT. Current code only checks for the upper bound, while the lower bound is unchecked. This can cause a lot of issues in the driver if the user passes 'db_count' as 0. Add a check for 0 also. While at it, remove the redundant 'db_count' variable from epf_ntb_configure_interrupt(). Fixes: 8b821cf76150 ("PCI: endpoint: Add EP function driver to provide NTB functionality") Signed-off-by: Manivannan Sadhasivam Signed-off-by: Manivannan Sadhasivam Signed-off-by: Bjorn Helgaas Link: https://patch.msgid.link/20260407124421.282766-3-mani@kernel.org Signed-off-by: Sasha Levin

PCI: endpoint: pci-epf-vntb: Add check to detect 'db_count' value of 0

2026-07-24T14:20:24+00:00

[ Upstream commit 33bd1ea748bc897c4d13437284e08c658e8d1340 ] epf_ntb->db_count value should be within 1 to MAX_DB_COUNT. Current code only checks for the upper bound, while the lower bound is unchecked. This can cause a lot of issues in the driver if the user passes 'db_count' as 0. Add a check for 0 also. While at it, remove the redundant 'db_count' assignment. Fixes: e35f56bb0330 ("PCI: endpoint: Support NTB transfer between RC and EP") Signed-off-by: Manivannan Sadhasivam Signed-off-by: Manivannan Sadhasivam Signed-off-by: Bjorn Helgaas Reviewed-by: Koichiro Den Reviewed-by: Frank Li Link: https://patch.msgid.link/20260407124421.282766-2-mani@kernel.org Signed-off-by: Sasha Levin

PCI: iproc: Restore .map_irq() for the platform bus driver

2026-07-24T14:20:18+00:00

[ Upstream commit 3c2e6cc6affa8acdb99a580be1f8f297edf54204 ] Commit b64aa11eb2dd ("PCI: Set bridge map_irq and swizzle_irq to default functions") moved the assignment of default .map_irq() callback to devm_of_pci_bridge_init() and removed the initialization of 'iproc_pcie::map_irq' in platform bus driver. This led to the callback getting assigned the NULL pointer for platform bus driver, thereby breaking the INTx functionality, since 'iproc_pcie::map_irq' overrides the 'pci_host_bridge::map_irq' callback in iproc_pcie_setup(). This issue only affected the iproc platform bus driver as this driver relies on the default callback for non-PAXC controllers. iproc-brcm driver was already providing the custom mapping function, so it was unaffected. Restore the original (and intended) behaviour to use the default map_irq function by removing the local 'iproc_pcie::map_irq' pointer and directly assigning the 'pci_host_bridge::map_irq' callback in iproc-bcma driver. This ensures that the default 'map_irq' callback is used for platform bus driver and only iproc-brcm driver overrides it with a custom one. Fixes: b64aa11eb2dd ("PCI: Set bridge map_irq and swizzle_irq to default functions") Signed-off-by: Mark Tomlinson [mani: commit log] Signed-off-by: Manivannan Sadhasivam Signed-off-by: Bjorn Helgaas Acked-by: Ray Jui Link: https://patch.msgid.link/20260430021628.1343154-1-mark.tomlinson@alliedtelesis.co.nz Signed-off-by: Sasha Levin

Revert "PCI/MSI: Unmap MSI-X region on error"

2026-07-24T14:20:12+00:00

[ Upstream commit f64e03da0d83cb173743888bff4a7e61476a8fc2 ] This reverts commit 1a8d4c6ecb4c81261bcdf13556abd4a958eca202. Commit 1a8d4c6ecb4c ("PCI/MSI: Unmap MSI-X region on error") added an iounmap(dev->msix_base) on the error path of msix_capability_init() to release the MSI-X region when msix_setup_interrupts() fails. When msix_setup_interrupts() fails, the call chain is: msix_setup_interrupts() -> __msix_setup_interrupts() struct pci_dev *dev __free(free_msi_irqs) = __dev; ... return ret; // __free cleanup fires on error The __free(free_msi_irqs) cleanup calls pci_free_msi_irqs(), which already handles the unmap: void pci_free_msi_irqs(struct pci_dev *dev) { pci_msi_teardown_msi_irqs(dev); if (dev->msix_base) { iounmap(dev->msix_base); // already unmapped here dev->msix_base = NULL; // and set to NULL } } So dev->msix_base is unmapped and set to NULL before msix_setup_interrupts() returns to msix_capability_init(). The "goto out_unmap" introduced by commit 1a8d4c6ecb4c ("PCI/MSI: Unmap MSI-X region on error") then calls iounmap() a second time on a NULL pointer. This was reproduced on Intel Emerald Rapids (192 CPUs) while running tools/testing/selftests/kexec/test_kexec_jump.sh: WARNING: CPU#44 at iounmap+0x2a/0xe0 RIP: 0010:iounmap+0x2a/0xe0 RDI: 0000000000000000 Call Trace: msix_capability_init+0x317/0x3f0 __pci_enable_msix_range+0x21d/0x2c0 pci_alloc_irq_vectors_affinity+0xa9/0x130 nvme_setup_io_queues+0x2a8/0x420 [nvme] nvme_reset_work+0x151/0x340 [nvme] ... RDI=0 confirms iounmap() is called with NULL. Restore the original "goto out_disable" and leave the unmap to the existing __free(free_msi_irqs) cleanup. Fixes: 1a8d4c6ecb4c ("PCI/MSI: Unmap MSI-X region on error") Reported-by: Guenter Roeck Signed-off-by: Yuanhe Shu Signed-off-by: Thomas Gleixner Link: https://lore.kernel.org/all/20260610194406.GA380991@bhelgaas/ Link: https://patch.msgid.link/20260611025901.1105209-1-xiangzao@linux.alibaba.com Closes: https://lore.kernel.org/all/4fc6208d-513b-4f41-a13a-4a0829ab50ad@roeck-us.net/ Signed-off-by: Sasha Levin