From: Laurentiu Tudor <laurentiu.tudor@nxp.com>
To: Fredrik Noring <noring@nocrew.org>
Cc: Robin Murphy <robin.murphy@arm.com>, "hch@lst.de" <hch@lst.de>,
"stern@rowland.harvard.edu" <stern@rowland.harvard.edu>,
"gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>,
"linux-usb@vger.kernel.org" <linux-usb@vger.kernel.org>,
"marex@denx.de" <marex@denx.de>,
"JuergenUrban@gmx.de" <JuergenUrban@gmx.de>,
Leo Li <leoyang.li@nxp.com>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [RFC PATCH v2 0/3] prerequisites for device reserved local mem rework
Date: Thu, 16 May 2019 09:35:27 +0000 [thread overview]
Message-ID: <0fea5fa7-e2b0-b34d-3bf8-976007df2dc2@nxp.com> (raw)
In-Reply-To: <20190515162858.GB17162@sx9>
Hi Fredrik,
Thanks very much for taking the time to look into this. Please see
comments inline.
On 15.05.2019 19:28, Fredrik Noring wrote:
> Hi Lauretniu,
>
>> I think that any recent kernel will do, so I'd say your current branch
>> should be fine.
>
> The kernel oopses with "unable to handle kernel paging request at virtual
> address 000aba0b" in hcd_alloc_coherent via usb_hcd_map_urb_for_dma.
By any chance, does this address looks like the dma_addr that the device
should target?
> This relates to patch 2/3 that I didn't quite understand, where
>
> - retval = dma_declare_coherent_memory(dev, mem->start,
> - mem->start - mem->parent->start,
> - resource_size(mem));
>
> becomes
>
> + retval = gen_pool_add_virt(hcd->localmem_pool,
> + (unsigned long)mem->start -
> + mem->parent->start,
> + mem->start, resource_size(mem),
>
> so that arguments two and three switch places. Given
>
> int dma_declare_coherent_memory(struct device *dev, phys_addr_t phys_addr,
> dma_addr_t device_addr, size_t size);
>
> and
>
> int gen_pool_add_virt(struct gen_pool *pool, unsigned long virt, phys_addr_t phys
> size_t size, int nid)
>
> it seems that the device address (dma_addr_t device_addr) now becomes a
> virtual address (unsigned long virt). Is that intended?
Actually, I think I'm misusing genalloc and also it appears that i need
to add a mapping on the phys address. So my plan is to change the
"unsigned long virt" to be the void * returned by the mapping operation
and the phys_addr_t be the dma_addr_t. I'll return with a patch.
Regarding the usage of unsigned long in genalloc, yeah seems a bit
strange but by looking at other users in the kernel it appears that
that's the design.
---
Best Regards, Laurentiu
> My corresponding patch is below for reference. I applied your 1/3 patch
> to test it.
>
> Fredrik
>
> diff --git a/drivers/usb/host/ohci-ps2.c b/drivers/usb/host/ohci-ps2.c
> --- a/drivers/usb/host/ohci-ps2.c
> +++ b/drivers/usb/host/ohci-ps2.c
> @@ -7,6 +7,7 @@
> */
>
> #include <linux/dma-mapping.h>
> +#include <linux/genalloc.h>
> #include <linux/module.h>
> #include <linux/platform_device.h>
> #include <linux/usb.h>
> @@ -92,12 +93,12 @@ static irqreturn_t ohci_ps2_irq(struct usb_hcd *hcd)
> return ohci_irq(hcd); /* Call normal IRQ handler. */
> }
>
> -static int iopheap_alloc_coherent(struct platform_device *pdev,
> - size_t size, int flags)
> +static int iopheap_alloc_coherent(struct platform_device *pdev, size_t size)
> {
> struct device *dev = &pdev->dev;
> struct usb_hcd *hcd = platform_get_drvdata(pdev);
> struct ps2_hcd *ps2priv = hcd_to_priv(hcd);
> + int err;
>
> if (ps2priv->iop_dma_addr != 0)
> return 0;
> @@ -112,28 +113,37 @@ static int iopheap_alloc_coherent(struct platform_device *pdev,
> return -ENOMEM;
> }
>
> - if (dma_declare_coherent_memory(dev,
> - iop_bus_to_phys(ps2priv->iop_dma_addr),
> - ps2priv->iop_dma_addr, size, flags)) {
> - dev_err(dev, "dma_declare_coherent_memory failed\n");
> - iop_free(ps2priv->iop_dma_addr);
> - ps2priv->iop_dma_addr = 0;
> - return -ENOMEM;
> + hcd->localmem_pool = devm_gen_pool_create(dev,
> + PAGE_SHIFT, dev_to_node(dev), DRV_NAME);
> + if (IS_ERR(hcd->localmem_pool)) {
> + err = PTR_ERR(hcd->localmem_pool);
> + goto out_err;
> + }
> +
> + err = gen_pool_add_virt(hcd->localmem_pool, ps2priv->iop_dma_addr,
> + iop_bus_to_phys(ps2priv->iop_dma_addr), size, dev_to_node(dev));
> + if (err < 0) {
> + dev_err(dev, "gen_pool_add_virt failed with %d\n", err);
> + goto out_err;
> }
>
> return 0;
> +
> +out_err:
> + iop_free(ps2priv->iop_dma_addr);
> + ps2priv->iop_dma_addr = 0;
> +
> + return err;
> }
>
> static void iopheap_free_coherent(struct platform_device *pdev)
> {
> - struct device *dev = &pdev->dev;
> struct usb_hcd *hcd = platform_get_drvdata(pdev);
> struct ps2_hcd *ps2priv = hcd_to_priv(hcd);
>
> if (ps2priv->iop_dma_addr == 0)
> return;
>
> - dma_release_declared_memory(dev);
> iop_free(ps2priv->iop_dma_addr);
> ps2priv->iop_dma_addr = 0;
> }
> @@ -177,8 +187,7 @@ static int ohci_hcd_ps2_probe(struct platform_device *pdev)
> goto err;
> }
>
> - ret = iopheap_alloc_coherent(pdev,
> - DMA_BUFFER_SIZE, DMA_MEMORY_EXCLUSIVE);
> + ret = iopheap_alloc_coherent(pdev, DMA_BUFFER_SIZE);
> if (ret != 0)
> goto err_alloc_coherent;
>
>
next prev parent reply other threads:[~2019-05-16 9:35 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-05-14 14:38 [RFC PATCH v2 0/3] prerequisites for device reserved local mem rework laurentiu.tudor
2019-05-14 14:38 ` [RFC PATCH v2 1/3] USB: use genalloc for USB HCs with local memory laurentiu.tudor
2019-05-14 14:42 ` Christoph Hellwig
2019-05-15 9:57 ` Laurentiu Tudor
2019-05-15 12:42 ` Christoph Hellwig
2019-05-14 14:38 ` [RFC PATCH v2 2/3] usb: host: ohci-sm501: init genalloc for " laurentiu.tudor
2019-05-14 14:38 ` [RFC PATCH v2 3/3] usb: host: ohci-tmio: " laurentiu.tudor
2019-05-14 15:17 ` [RFC PATCH v2 0/3] prerequisites for device reserved local mem rework Robin Murphy
2019-05-14 18:29 ` Fredrik Noring
2019-05-15 10:37 ` Laurentiu Tudor
2019-05-15 16:28 ` Fredrik Noring
2019-05-16 9:35 ` Laurentiu Tudor [this message]
2019-05-16 13:47 ` Fredrik Noring
2019-05-16 11:49 ` Laurentiu Tudor
2019-05-16 15:15 ` Fredrik Noring
2019-05-17 10:52 ` Laurentiu Tudor
2019-05-17 17:41 ` Fredrik Noring
2019-05-20 11:34 ` Laurentiu Tudor
2019-05-20 15:41 ` Fredrik Noring
2019-05-21 10:32 ` hch
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=0fea5fa7-e2b0-b34d-3bf8-976007df2dc2@nxp.com \
--to=laurentiu.tudor@nxp.com \
--cc=JuergenUrban@gmx.de \
--cc=gregkh@linuxfoundation.org \
--cc=hch@lst.de \
--cc=leoyang.li@nxp.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-usb@vger.kernel.org \
--cc=marex@denx.de \
--cc=noring@nocrew.org \
--cc=robin.murphy@arm.com \
--cc=stern@rowland.harvard.edu \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).