kernel/drivers/pci/switch/switchtec.c, branch linux-5.1.y

switchtec: Fix unintended mask of MRPC event

2019-06-15T09:53:00Z

[ Upstream commit 083c1b5e50b701899dc32445efa8b153685260d5 ] When running application tool switchtec-user's `firmware update` and `event wait` commands concurrently, sometimes the firmware update speed reduced significantly. It is because when the MRPC event happened after MRPC event occurrence check but before the event mask loop reaches its header register in event ISR, the MRPC event would be masked unintentionally. Since there's no chance to enable it again except for a module reload, all the following MRPC execution completion checks time out. Fix this bug by skipping the mask operation for MRPC event in event ISR, same as what we already do for LINK event. Fixes: 52eabba5bcdb ("switchtec: Add IOCTLs to the Switchtec driver") Signed-off-by: Wesley Sheng Signed-off-by: Bjorn Helgaas Reviewed-by: Logan Gunthorpe Signed-off-by: Sasha Levin

cross-tree: phase out dma_zalloc_coherent()

2019-01-08T12:58:37Z

We already need to zero out memory for dma_alloc_coherent(), as such using dma_zalloc_coherent() is superflous. Phase it out. This change was generated with the following Coccinelle SmPL patch: @ replace_dma_zalloc_coherent @ expression dev, size, data, handle, flags; @@ -dma_zalloc_coherent(dev, size, handle, flags) +dma_alloc_coherent(dev, size, handle, flags) Suggested-by: Christoph Hellwig Signed-off-by: Luis Chamberlain [hch: re-ran the script on the latest tree] Signed-off-by: Christoph Hellwig

switchtec: Add MRPC DMA mode support

2018-12-13T15:03:32Z

MRPC normal mode requires the host to read the MRPC command status and output data from BAR. This results in high latency responses from the Memory Read TLP and potential Completion Timeout (CTO). Add support for MRPC DMA mode, including related macro definitions and data structures and code to: * Retrieve MRPC DMA mode version from adapter firmware * Allocate DMA buffer, register ISR, and enable DMA during init * Check MRPC execution status and get execution results from DMA buffer * Release DMA buffer and disable DMA function when unloading module MRPC DMA mode is a new feature of firmware, and the driver will fall back to MRPC normal mode if there is no support in the legacy firmware. Add a module parameter, "use_dma_mrpc", to select between MRPC DMA mode and MRPC normal mode. Since the driver automatically detects DMA support in the firmware, this parameter is just for debugging and testing. Include so that readq/writeq is replaced by two readl/writel on systems that do not support it. Signed-off-by: Wesley Sheng [bhelgaas: changelog, simplify dma_ver check] Signed-off-by: Bjorn Helgaas Reviewed-by: Logan Gunthorpe

switchtec: Improve MRPC efficiency by enabling write combining

2018-12-13T15:03:24Z

The MRPC Input buffer is mostly memory without any side effects, so we can improve the access time by enabling write combining on this region of the BAR. In a few places, we still need to flush the WC buffer. To do this, we simply read from the Outbound Doorbell register because reads to this register are processed by low latency hardware. Signed-off-by: Kelvin Cao Signed-off-by: Wesley Sheng Signed-off-by: Bjorn Helgaas Reviewed-by: Logan Gunthorpe

switchtec: Fix SWITCHTEC_IOCTL_EVENT_IDX_ALL flags overwrite

2018-12-12T22:32:16Z

In the ioctl_event_ctl() SWITCHTEC_IOCTL_EVENT_IDX_ALL case, we call event_ctl() several times with the same "ctl" struct. Each call clobbers ctl.flags, which leads to the problem that we may not actually enable or disable all events as the user requested. Preserve the event flag value with a temporary variable. Fixes: 52eabba5bcdb ("switchtec: Add IOCTLs to the Switchtec driver") Signed-off-by: Joey Zhang Signed-off-by: Wesley Sheng [bhelgaas: changelog] Signed-off-by: Bjorn Helgaas Reviewed-by: Logan Gunthorpe

switchtec: Set DMA coherent mask

2018-12-12T22:31:54Z

Switchtec hardware supports 64-bit DMA, so set the correct DMA mask. This allows the CMA to allocate larger buffers for memory windows. Signed-off-by: Boris Glimcher Signed-off-by: Wesley Sheng Signed-off-by: Bjorn Helgaas Reviewed-by: Logan Gunthorpe

switchtec: Remove immediate status check after submitting MRPC command

2018-12-12T22:30:03Z

After submitting a Firmware Download MRPC command, Switchtec firmware will delay Management EP BAR MemRd TLP responses by more than 10ms. This is a firmware limitation. Delayed MemRd completions are a problem for systems with a low Completion Timeout (CTO). The current driver checks the MRPC status immediately after submitting an MRPC command, which results in a delayed MemRd completion that may cause a Completion Timeout. Remove the immediate status check and rely on the check after receiving an interrupt or timing out. This is only a software workaround to the READ issue and a proper fix of this should be done in firmware. Fixes: 080b47def5e5 ("MicroSemi Switchtec management interface driver") Signed-off-by: Kelvin Cao Signed-off-by: Wesley Sheng Signed-off-by: Bjorn Helgaas Reviewed-by: Logan Gunthorpe

switchtec: Fix Spectre v1 vulnerability

2018-09-11T13:47:40Z

p.port can is indirectly controlled by user-space, hence leading to a potential exploitation of the Spectre variant 1 vulnerability. This issue was detected with the help of Smatch: drivers/pci/switch/switchtec.c:912 ioctl_port_to_pff() warn: potential spectre issue 'pcfg->dsp_pff_inst_id' [r] Fix this by sanitizing p.port before using it to index pcfg->dsp_pff_inst_id Notice that given that speculation windows are large, the policy is to kill the speculation on the first load and not worry if it can be completed with a dependent load/store [1]. [1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2 Signed-off-by: Gustavo A. R. Silva Signed-off-by: Bjorn Helgaas Acked-by: Logan Gunthorpe Cc: stable@vger.kernel.org

switchtec: Use generic PCI Vendor ID and Class Code

2018-06-30T01:16:44Z

Move the Microsemi Switchtec PCI Vendor ID (same as PCI_VENDOR_ID_PMC_Sierra) to pci_ids.h. Also, replace Microsemi class constants with the standard PCI definitions. Signed-off-by: Doug Meyer [bhelgaas: restore SPDX (I assume it was removed by mistake), remove device ID definitions] Signed-off-by: Bjorn Helgaas Reviewed-by: Logan Gunthorpe

vfs: do bulk POLL* -> EPOLL* replacement

2018-02-11T22:34:03Z

This is the mindless scripted replacement of kernel use of POLL* variables as described by Al, done by this script: for V in IN OUT PRI ERR RDNORM RDBAND WRNORM WRBAND HUP RDHUP NVAL MSG; do L=`git grep -l -w POLL$V | grep -v '^t' | grep -v /um/ | grep -v '^sa' | grep -v '/poll.h$'|grep -v '^D'` for f in $L; do sed -i "-es/^$[^\"]*$$\$/\\1E\\2/" $f; done done with de-mangling cleanups yet to come. NOTE! On almost all architectures, the EPOLL* constants have the same values as the POLL* constants do. But they keyword here is "almost". For various bad reasons they aren't the same, and epoll() doesn't actually work quite correctly in some cases due to this on Sparc et al. The next patch from Al will sort out the final differences, and we should be all done. Scripted-by: Al Viro Signed-off-by: Linus Torvalds