On Wed, Sep 14, 2016 at 10:38:55AM -0600, Jon Derrick wrote:
> AER injecting tests with many devices and nosourceid are resulting in
> soft lockups, mostly due to the config reads (which have been helped
> by recent aer capability position caching patches), but also due to a
> superfluous kmalloc/kfree pair executed for each p_device.
>
> When a device emits an error notification, it's not unreasonable to
> assume that it may emit another error notification soon. Instead of
> kmallocing and kfreeing the aer_err_info struct for each p_device during
> an aer_isr pass, move the struct into the aer root port container and
> kill the allocation.
>
> The current code suggests the aer_err_info struct could be huge, but is
> actually only about 70 bytes. Whereas the aer_rpc root port container is
> over 900 bytes. Because of the already large size of the root port
> container, moving the struct into it should not adversely impact access
> to the root port container.
>
> Suggested-by: Bjorn Helgaas <helgaas@xxxxxxxxxx>
> Signed-off-by: Jon Derrick <jonathan.derrick@xxxxxxxxx>
Applied to pci/aer for v4.9, thanks, Jon!
> ---
> drivers/pci/pcie/aer/aerdrv.h | 1 +
> drivers/pci/pcie/aer/aerdrv_core.c | 13 ++-----------
> 2 files changed, 3 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/pci/pcie/aer/aerdrv.h b/drivers/pci/pcie/aer/aerdrv.h
> index f15ca8d..d51e4a5 100644
> --- a/drivers/pci/pcie/aer/aerdrv.h
> +++ b/drivers/pci/pcie/aer/aerdrv.h
> @@ -60,6 +60,7 @@ struct aer_rpc {
> struct pcie_device *rpd; /* Root Port device */
> struct work_struct dpc_handler;
> struct aer_err_source e_sources[AER_ERROR_SOURCES_MAX];
> + struct aer_err_info e_info;
> unsigned short prod_idx; /* Error Producer Index */
> unsigned short cons_idx; /* Error Consumer Index */
> int isr;
> diff --git a/drivers/pci/pcie/aer/aerdrv_core.c b/drivers/pci/pcie/aer/aerdrv_core.c
> index 8262527..9fd18a0 100644
> --- a/drivers/pci/pcie/aer/aerdrv_core.c
> +++ b/drivers/pci/pcie/aer/aerdrv_core.c
> @@ -711,15 +711,8 @@ static inline void aer_process_err_devices(struct pcie_device *p_device,
> static void aer_isr_one_error(struct pcie_device *p_device,
> struct aer_err_source *e_src)
> {
> - struct aer_err_info *e_info;
> -
> - /* struct aer_err_info might be big, so we allocate it with slab */
> - e_info = kmalloc(sizeof(struct aer_err_info), GFP_KERNEL);
> - if (!e_info) {
> - dev_printk(KERN_DEBUG, &p_device->port->dev,
> - "Can't allocate mem when processing AER errors\n");
> - return;
> - }
> + struct aer_rpc *rpc = get_service_data(p_device);
> + struct aer_err_info *e_info = &rpc->e_info;
>
> /*
> * There is a possibility that both correctable error and
> @@ -758,8 +751,6 @@ static void aer_isr_one_error(struct pcie_device *p_device,
> if (find_source_device(p_device->port, e_info))
> aer_process_err_devices(p_device, e_info);
> }
> -
> - kfree(e_info);
> }
>
> /**
> --
> 1.8.3.1
>
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
