[PATCH 2/2] PCI: iproc: add bcma pcie driver

On Wed, May 13, 2015 at 08:43:47PM +0200, Hauke Mehrtens wrote:

On 05/13/2015 06:30 PM, Ray Jui wrote:

quoted

Hi Rafal,

On 5/13/2015 9:19 AM, Rafa? Mi?ecki wrote:

quoted

On 13 May 2015 at 17:56, Ray Jui [off-list ref] wrote:

quoted

On 5/12/2015 11:27 PM, Rafa? Mi?ecki wrote:

quoted

On 12 May 2015 at 23:23, Hauke Mehrtens [off-list ref] wrote:

quoted

This driver adds support for the PCIe 2.0 controller found on the bcma
bus. This controller can be found on (mostly) all Broadcom BCM470X /
BCM5301X ARM SoCs.

The driver found in the Broadcom SDK does some more stuff, like setting
up some DMA memory areas, chaining MPS and MRRS to 512 and also some
PHY changes like "improving" the PCIe jitter and doing some special
initializations for the 3rd PCIe port.

This was tested on a bcm4708 board with 2 PCIe ports and wireless cards
connected to them.

PCI_DOMAINS is needed by this driver, because normally there is more
than one PCIe controller and without PCI_DOMAINS only the first
controller gets registered.
This controller gets 6 IRQs, the last one is trigged by all IRQ events.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>

Acked-by: Rafa? Mi?ecki <zajec5@gmail.com>

quoted

+static int iproc_pcie_bcma_probe(struct bcma_device *bdev)
+{
+       struct iproc_pcie *pcie;
+       LIST_HEAD(res);
+       struct resource res_mem;
+       int ret;
+
+       pcie = devm_kzalloc(&bdev->dev, sizeof(*pcie), GFP_KERNEL);
+       if (!pcie)
+               return -ENOMEM;
+
+       pcie->dev = &bdev->dev;
+       bcma_set_drvdata(bdev, pcie);
+
+       pcie->base = bdev->io_addr;
+
+       res_mem.start = bdev->addr_s[0];
+       res_mem.end = bdev->addr_s[0] + SZ_128M - 1;
+       res_mem.name = "PCIe MEM space";
+       res_mem.flags = IORESOURCE_MEM;
+       pci_add_resource(&res, &res_mem);
+
+       pcie->resources = &res;
+
+       pcie->map_irq = iproc_pcie_bcma_map_irq;
+
+       ret = iproc_pcie_setup(pcie);

I think I don't like this part of iproc design. It lefts
pcie->resources pointing to some random memory after the setup/probe
are done. Guess it should be a separated parameter or sth.

The patch is still OK, I just refer to generic iproc possible issue.

Sorry Rafal, but could you please be more specific on this?

iproc_pcie_pltfm_probe (and iproc_pcie_bcma_probe) have local "res"
variable (each its own). They do:
pcie->resources = &res;
and then they call
iproc_pcie_setup(pcie);

After iproc_pcie_pltfm_probe / iproc_pcie_bcma_probe returns, the
pointer pcie->resources is not valid anymore, yet pcie struct is still
in use. Of course pcie->resources isn't used anymore, but still, it's
some in-struct pointer (to the random memory since local variable is
not accessible anymore).

I think you should drop
struct list_head *resources;
from the struct iproc_pcie and use
iproc_pcie_setup(struct iproc_pcie *pcie, struct list_head *resources)

Okay thanks. That makes sense.

Or I should keep a copy of the resources under pcie->resources. In the
current pcie-iproc.c, the resource is not used anywhere else except when
creating the root bus under iproc_pcie_setup. But in the future, I might
need to add more code to explicitly program the outbound/inbound windows
so this driver can work with some other iProc SoCs where the desired
windows do not match power-on-reset values.

I plan to change this along with the window programming patch in the
future. I also think Hauke's current patch is fine.

Thanks,

Ray

I think parts of this resources are allocated and never freed.

pci_add_resource() allocates some bytes, but I do not see where they are
freed, this also applies to the platform driver.

Where are we on this patch?  I see Ray's ack for [1/2], and a "I think the
current patch is fine"; is that an ack for [2/2] as well?

Bjorn