Propper use of pci_map_sg

Malte Vesper <malte.vesper@xxxxxxxxxxxxxxxxxxxxxxxxx> · Thu, 29 Jan 2015 19:50:52 +0000



    Hi,

    I am trying to use pci_map_sg in combination with
    get_user_pages_fast for a driver. However my code screws the
    processes memory map over. For easier testing I bundled the
    piinin/unpinning and mapping unmapping into one ioctl call. The
    following code misses all the checks on return values, since it is
    only inteneded as a MWE.

    
    When I run my code I get a "BUG: Bad page map in process ...",
    after "Done" is printed.

    
    I tried following DMA-API-howto.txt and looking at other code, but I
    fail to see where I go wrong.

    
    Regards, Ted

    
    Code from the IOCTL handler:

    
            case IOCTL_FPGA_PIN_PAGE:

            {

                struct pageInfo pageInfo;

    
                dev_dbg(&pcidev->dev, pr_fmt("IOCTL:
    IOCTL_FPGA_PIN_PAGE\n"));

    
                if(!copy_from_user(&pageInfo, (void*)arg,
    sizeof(struct pageInfo))) {

                    //horrible test

                    const int noPages = pageInfo.size/PAGE_SIZE;

                    int pinned;

                    int mapped  = 0;

                    struct scatterlist* scatterlist;

                    printk("Test start\n");

                    //userspacestartpointer, nopages, write?, page*
    array

                    struct page** pages=kmalloc(sizeof(struct
    page*)*noPages, GFP_KERNEL);

                    pinned=get_user_pages_fast((unsigned
    long)pageInfo.start, noPages, 1, pages);

    
                    scatterlist = kmalloc(sizeof(struct
    scatterlist)*pinned, GFP_KERNEL);

    
                    for(int i=0; i<pinned; ++i) {

                        sg_set_page(&scatterlist[i], pages[i],
    PAGE_SIZE, 0);

                    }

    
                    mapped = pci_map_sg(pcidev, scatterlist,
    pinned, DMA_BIDIRECTIONAL);

                    pci_unmap_sg(pcidev, scatterlist, pinned,
    DMA_BIDIRECTIONAL);

    
                    for(int i=0; i<pinned; ++i) {

                        put_page(pages[i]); //I did place a
    print here and got two pages unpinned as I expected

                    }

                    kfree(scatterlist);

                    kfree(pages);

                    printk("Done\n");

                    /*pageInfo.pageId =
    getPageId(getArea(&pageInfo));

                    copy_to_user((void*)arg, &pageInfo,
    sizeof(struct pageInfo));*/

                    return pageInfo.pageId;

                } else {

                    return -EFAULT;

                }

            }

                break;

    
    P.S.: I used the following code to allocate the memory in userspace,
    and I use chunk and chunksize to fill pageInfo.start and
    pageInfo.size respectivly:

    
        const int pagesize = sysconf(_SC_PAGESIZE);

        const int chunksize = 2*pagesize;

        void* chunk;

    
        if(int error = posix_memalign(&chunk, pagesize, chunksize))
    {

            std::cout << "Could not get a page." <<
    std::endl;

            return error;

        }

  
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@xxxxxxxxxxxxxxxxx
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies