ipath_file_ops.c

来自「linux 内核源代码」· C语言 代码 · 共 2,267 行 · 第 1/5 页

			gfp_flags);

		if (!pd->port_rcvegrbuf[e]) {
			ret = -ENOMEM;
			goto bail_rcvegrbuf_phys;
		}
	}

	pd->port_rcvegr_phys = pd->port_rcvegrbuf_phys[0];

	for (e = chunk = 0; chunk < pd->port_rcvegrbuf_chunks; chunk++) {
		dma_addr_t pa = pd->port_rcvegrbuf_phys[chunk];
		unsigned i;

		for (i = 0; e < egrcnt && i < egrperchunk; e++, i++) {
			dd->ipath_f_put_tid(dd, e + egroff +
					    (u64 __iomem *)
					    ((char __iomem *)
					     dd->ipath_kregbase +
					     dd->ipath_rcvegrbase),
					    RCVHQ_RCV_TYPE_EAGER, pa);
			pa += egrsize;
		}
		cond_resched();	/* don't hog the cpu */
	}

	ret = 0;
	goto bail;

bail_rcvegrbuf_phys:
	for (e = 0; e < pd->port_rcvegrbuf_chunks &&
		pd->port_rcvegrbuf[e]; e++) {
		dma_free_coherent(&dd->pcidev->dev, size,
				  pd->port_rcvegrbuf[e],
				  pd->port_rcvegrbuf_phys[e]);

	}
	kfree(pd->port_rcvegrbuf_phys);
	pd->port_rcvegrbuf_phys = NULL;
bail_rcvegrbuf:
	kfree(pd->port_rcvegrbuf);
	pd->port_rcvegrbuf = NULL;
bail:
	return ret;
}


/* common code for the mappings on dma_alloc_coherent mem */
static int ipath_mmap_mem(struct vm_area_struct *vma,
	struct ipath_portdata *pd, unsigned len, int write_ok,
	void *kvaddr, char *what)
{
	struct ipath_devdata *dd = pd->port_dd;
	unsigned long pfn;
	int ret;

	if ((vma->vm_end - vma->vm_start) > len) {
		dev_info(&dd->pcidev->dev,
		         "FAIL on %s: len %lx > %x\n", what,
			 vma->vm_end - vma->vm_start, len);
		ret = -EFAULT;
		goto bail;
	}

	if (!write_ok) {
		if (vma->vm_flags & VM_WRITE) {
			dev_info(&dd->pcidev->dev,
				 "%s must be mapped readonly\n", what);
			ret = -EPERM;
			goto bail;
		}

		/* don't allow them to later change with mprotect */
		vma->vm_flags &= ~VM_MAYWRITE;
	}

	pfn = virt_to_phys(kvaddr) >> PAGE_SHIFT;
	ret = remap_pfn_range(vma, vma->vm_start, pfn,
			      len, vma->vm_page_prot);
	if (ret)
		dev_info(&dd->pcidev->dev, "%s port%u mmap of %lx, %x "
			 "bytes r%c failed: %d\n", what, pd->port_port,
			 pfn, len, write_ok?'w':'o', ret);
	else
		ipath_cdbg(VERBOSE, "%s port%u mmaped %lx, %x bytes "
			   "r%c\n", what, pd->port_port, pfn, len,
			   write_ok?'w':'o');
bail:
	return ret;
}

static int mmap_ureg(struct vm_area_struct *vma, struct ipath_devdata *dd,
		     u64 ureg)
{
	unsigned long phys;
	int ret;

	/*
	 * This is real hardware, so use io_remap.  This is the mechanism
	 * for the user process to update the head registers for their port
	 * in the chip.
	 */
	if ((vma->vm_end - vma->vm_start) > PAGE_SIZE) {
		dev_info(&dd->pcidev->dev, "FAIL mmap userreg: reqlen "
			 "%lx > PAGE\n", vma->vm_end - vma->vm_start);
		ret = -EFAULT;
	} else {
		phys = dd->ipath_physaddr + ureg;
		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);

		vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND;
		ret = io_remap_pfn_range(vma, vma->vm_start,
					 phys >> PAGE_SHIFT,
					 vma->vm_end - vma->vm_start,
					 vma->vm_page_prot);
	}
	return ret;
}

static int mmap_piobufs(struct vm_area_struct *vma,
			struct ipath_devdata *dd,
			struct ipath_portdata *pd,
			unsigned piobufs, unsigned piocnt)
{
	unsigned long phys;
	int ret;

	/*
	 * When we map the PIO buffers in the chip, we want to map them as
	 * writeonly, no read possible.   This prevents access to previous
	 * process data, and catches users who might try to read the i/o
	 * space due to a bug.
	 */
	if ((vma->vm_end - vma->vm_start) > (piocnt * dd->ipath_palign)) {
		dev_info(&dd->pcidev->dev, "FAIL mmap piobufs: "
			 "reqlen %lx > PAGE\n",
			 vma->vm_end - vma->vm_start);
		ret = -EINVAL;
		goto bail;
	}

	phys = dd->ipath_physaddr + piobufs;

	/*
	 * Don't mark this as non-cached, or we don't get the
	 * write combining behavior we want on the PIO buffers!
	 */

#if defined(__powerpc__)
	/* There isn't a generic way to specify writethrough mappings */
	pgprot_val(vma->vm_page_prot) |= _PAGE_NO_CACHE;
	pgprot_val(vma->vm_page_prot) |= _PAGE_WRITETHRU;
	pgprot_val(vma->vm_page_prot) &= ~_PAGE_GUARDED;
#endif

	/*
	 * don't allow them to later change to readable with mprotect (for when
	 * not initially mapped readable, as is normally the case)
	 */
	vma->vm_flags &= ~VM_MAYREAD;
	vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND;

	ret = io_remap_pfn_range(vma, vma->vm_start, phys >> PAGE_SHIFT,
				 vma->vm_end - vma->vm_start,
				 vma->vm_page_prot);
bail:
	return ret;
}

static int mmap_rcvegrbufs(struct vm_area_struct *vma,
			   struct ipath_portdata *pd)
{
	struct ipath_devdata *dd = pd->port_dd;
	unsigned long start, size;
	size_t total_size, i;
	unsigned long pfn;
	int ret;

	size = pd->port_rcvegrbuf_size;
	total_size = pd->port_rcvegrbuf_chunks * size;
	if ((vma->vm_end - vma->vm_start) > total_size) {
		dev_info(&dd->pcidev->dev, "FAIL on egr bufs: "
			 "reqlen %lx > actual %lx\n",
			 vma->vm_end - vma->vm_start,
			 (unsigned long) total_size);
		ret = -EINVAL;
		goto bail;
	}

	if (vma->vm_flags & VM_WRITE) {
		dev_info(&dd->pcidev->dev, "Can't map eager buffers as "
			 "writable (flags=%lx)\n", vma->vm_flags);
		ret = -EPERM;
		goto bail;
	}
	/* don't allow them to later change to writeable with mprotect */
	vma->vm_flags &= ~VM_MAYWRITE;

	start = vma->vm_start;

	for (i = 0; i < pd->port_rcvegrbuf_chunks; i++, start += size) {
		pfn = virt_to_phys(pd->port_rcvegrbuf[i]) >> PAGE_SHIFT;
		ret = remap_pfn_range(vma, start, pfn, size,
				      vma->vm_page_prot);
		if (ret < 0)
			goto bail;
	}
	ret = 0;

bail:
	return ret;
}

/*
 * ipath_file_vma_nopage - handle a VMA page fault.
 */
static struct page *ipath_file_vma_nopage(struct vm_area_struct *vma,
					  unsigned long address, int *type)
{
	unsigned long offset = address - vma->vm_start;
	struct page *page = NOPAGE_SIGBUS;
	void *pageptr;

	/*
	 * Convert the vmalloc address into a struct page.
	 */
	pageptr = (void *)(offset + (vma->vm_pgoff << PAGE_SHIFT));
	page = vmalloc_to_page(pageptr);
	if (!page)
		goto out;

	/* Increment the reference count. */
	get_page(page);
	if (type)
		*type = VM_FAULT_MINOR;
out:
	return page;
}

static struct vm_operations_struct ipath_file_vm_ops = {
	.nopage = ipath_file_vma_nopage,
};

static int mmap_kvaddr(struct vm_area_struct *vma, u64 pgaddr,
		       struct ipath_portdata *pd, unsigned subport)
{
	unsigned long len;
	struct ipath_devdata *dd;
	void *addr;
	size_t size;
	int ret = 0;

	/* If the port is not shared, all addresses should be physical */
	if (!pd->port_subport_cnt)
		goto bail;

	dd = pd->port_dd;
	size = pd->port_rcvegrbuf_chunks * pd->port_rcvegrbuf_size;

	/*
	 * Each process has all the subport uregbase, rcvhdrq, and
	 * rcvegrbufs mmapped - as an array for all the processes,
	 * and also separately for this process.
	 */
	if (pgaddr == cvt_kvaddr(pd->subport_uregbase)) {
		addr = pd->subport_uregbase;
		size = PAGE_SIZE * pd->port_subport_cnt;
	} else if (pgaddr == cvt_kvaddr(pd->subport_rcvhdr_base)) {
		addr = pd->subport_rcvhdr_base;
		size = pd->port_rcvhdrq_size * pd->port_subport_cnt;
	} else if (pgaddr == cvt_kvaddr(pd->subport_rcvegrbuf)) {
		addr = pd->subport_rcvegrbuf;
		size *= pd->port_subport_cnt;
        } else if (pgaddr == cvt_kvaddr(pd->subport_uregbase +
                                        PAGE_SIZE * subport)) {
                addr = pd->subport_uregbase + PAGE_SIZE * subport;
                size = PAGE_SIZE;
        } else if (pgaddr == cvt_kvaddr(pd->subport_rcvhdr_base +
                                pd->port_rcvhdrq_size * subport)) {
                addr = pd->subport_rcvhdr_base +
                        pd->port_rcvhdrq_size * subport;
                size = pd->port_rcvhdrq_size;
        } else if (pgaddr == cvt_kvaddr(pd->subport_rcvegrbuf +
                               size * subport)) {
                addr = pd->subport_rcvegrbuf + size * subport;
                /* rcvegrbufs are read-only on the slave */
                if (vma->vm_flags & VM_WRITE) {
                        dev_info(&dd->pcidev->dev,
                                 "Can't map eager buffers as "
                                 "writable (flags=%lx)\n", vma->vm_flags);
                        ret = -EPERM;
                        goto bail;
                }
                /*
                 * Don't allow permission to later change to writeable
                 * with mprotect.
                 */
                vma->vm_flags &= ~VM_MAYWRITE;
	} else {
		goto bail;
	}
	len = vma->vm_end - vma->vm_start;
	if (len > size) {
		ipath_cdbg(MM, "FAIL: reqlen %lx > %zx\n", len, size);
		ret = -EINVAL;
		goto bail;
	}

	vma->vm_pgoff = (unsigned long) addr >> PAGE_SHIFT;
	vma->vm_ops = &ipath_file_vm_ops;
	vma->vm_flags |= VM_RESERVED | VM_DONTEXPAND;
	ret = 1;

bail:
	return ret;
}

/**
 * ipath_mmap - mmap various structures into user space
 * @fp: the file pointer
 * @vma: the VM area
 *
 * We use this to have a shared buffer between the kernel and the user code
 * for the rcvhdr queue, egr buffers, and the per-port user regs and pio
 * buffers in the chip.  We have the open and close entries so we can bump
 * the ref count and keep the driver from being unloaded while still mapped.
 */
static int ipath_mmap(struct file *fp, struct vm_area_struct *vma)
{
	struct ipath_portdata *pd;
	struct ipath_devdata *dd;
	u64 pgaddr, ureg;
	unsigned piobufs, piocnt;
	int ret;

	pd = port_fp(fp);
	if (!pd) {
		ret = -EINVAL;
		goto bail;
	}
	dd = pd->port_dd;

	/*
	 * This is the ipath_do_user_init() code, mapping the shared buffers
	 * into the user process. The address referred to by vm_pgoff is the
	 * file offset passed via mmap().  For shared ports, this is the
	 * kernel vmalloc() address of the pages to share with the master.
	 * For non-shared or master ports, this is a physical address.
	 * We only do one mmap for each space mapped.
	 */
	pgaddr = vma->vm_pgoff << PAGE_SHIFT;

	/*
	 * Check for 0 in case one of the allocations failed, but user
	 * called mmap anyway.
	 */
	if (!pgaddr)  {
		ret = -EINVAL;
		goto bail;
	}

	ipath_cdbg(MM, "pgaddr %llx vm_start=%lx len %lx port %u:%u:%u\n",
		   (unsigned long long) pgaddr, vma->vm_start,
		   vma->vm_end - vma->vm_start, dd->ipath_unit,
		   pd->port_port, subport_fp(fp));

	/*
	 * Physical addresses must fit in 40 bits for our hardware.
	 * Check for kernel virtual addresses first, anything else must
	 * match a HW or memory address.
	 */
	ret = mmap_kvaddr(vma, pgaddr, pd, subport_fp(fp));
	if (ret) {
		if (ret > 0)
			ret = 0;
		goto bail;
	}

	ureg = dd->ipath_uregbase + dd->ipath_palign * pd->port_port;
	if (!pd->port_subport_cnt) {
		/* port is not shared */
		piocnt = dd->ipath_pbufsport;
		piobufs = pd->port_piobufs;
	} else if (!subport_fp(fp)) {
		/* caller is the master */
		piocnt = (dd->ipath_pbufsport / pd->port_subport_cnt) +
			 (dd->ipath_pbufsport % pd->port_subport_cnt);
		piobufs = pd->port_piobufs +
			dd->ipath_palign * (dd->ipath_pbufsport - piocnt);
	} else {
		unsigned slave = subport_fp(fp) - 1;

		/* caller is a slave */
		piocnt = dd->ipath_pbufsport / pd->port_subport_cnt;
		piobufs = pd->port_piobufs + dd->ipath_palign * piocnt * slave;
	}

	if (pgaddr == ureg)
		ret = mmap_ureg(vma, dd, ureg);
	else if (pgaddr == piobufs)
		ret = mmap_piobufs(vma, dd, pd, piobufs, piocnt);
	else if (pgaddr == dd->ipath_pioavailregs_phys)
		/* in-memory copy of pioavail registers */
		ret = ipath_mmap_mem(vma, pd, PAGE_SIZE, 0,
			      	     (void *) dd->ipath_pioavailregs_dma,
				     "pioavail registers");
	else if (pgaddr == pd->port_rcvegr_phys)
		ret = mmap_rcvegrbufs(vma, pd);
	else if (pgaddr == (u64) pd->port_rcvhdrq_phys)
		/*
		 * The rcvhdrq itself; readonly except on HT (so have
		 * to allow writable mapping), multiple pages, contiguous
		 * from an i/o perspective.
		 */
		ret = ipath_mmap_mem(vma, pd, pd->port_rcvhdrq_size, 1,
				     pd->port_rcvhdrq,
				     "rcvhdrq");
	else if (pgaddr == (u64) pd->port_rcvhdrqtailaddr_phys)
		/* in-memory copy of rcvhdrq tail register */
		ret = ipath_mmap_mem(vma, pd, PAGE_SIZE, 0,
				     pd->port_rcvhdrtail_kvaddr,
				     "rcvhdrq tail");
	else
		ret = -EINVAL;

	vma->vm_private_data = NULL;

	if (ret < 0)
		dev_info(&dd->pcidev->dev,
			 "Failure %d on off %llx len %lx\n",
			 -ret, (unsigned long long)pgaddr,
			 vma->vm_end - vma->vm_start);
bail:
	return ret;
}

static unsigned ipath_poll_hdrqfull(struct ipath_portdata *pd)
{
	unsigned pollflag = 0;

	if ((pd->poll_type & IPATH_POLL_TYPE_OVERFLOW) &&
	    pd->port_hdrqfull != pd->port_hdrqfull_poll) {
		pollflag |= POLLIN | POLLRDNORM;
		pd->port_hdrqfull_poll = pd->port_hdrqfull;
	}

	return pollflag;
}

static unsigned int ipath_poll_urgent(struct ipath_portdata *pd,
				      struct file *fp,
				      struct poll_table_struct *pt)
{
	unsigned pollflag = 0;