[PATCH 04/16] ehca: userspace support

[PATCH 04/16] ehca: userspace support

[Heiko J Schick]

Signed-off-by: Heiko J Schick <schickhj@de.ibm.com>


  ehca_uverbs.c |  409 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
  1 file changed, 409 insertions(+)



--- linux-2.6.17-rc2-orig/drivers/infiniband/hw/ehca/ehca_uverbs.c	1970-01-01 01:00:00.000000000 +0100
+++ linux-2.6.17-rc2/drivers/infiniband/hw/ehca/ehca_uverbs.c	2006-04-24 15:12:03.000000000 +0200
@@ -0,0 +1,409 @@
+/*
+ *  IBM eServer eHCA Infiniband device driver for Linux on POWER
+ *
+ *  userspace support verbs
+ *
+ *  Authors: Christoph Raisch <raisch@de.ibm.com>
+ *           Hoang-Nam Nguyen <hnguyen@de.ibm.com>
+ *           Heiko J Schick <schickhj@de.ibm.com>
+ *
+ *  Copyright (c) 2005 IBM Corporation
+ *
+ *  All rights reserved.
+ *
+ *  This source code is distributed under a dual license of GPL v2.0 and OpenIB
+ *  BSD.
+ *
+ * OpenIB BSD License
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials
+ * provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
+ * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER
+ * IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ *
+ *  $Id: ehca_uverbs.c,v 1.18 2006/04/24 13:12:03 schickhj Exp $
+ */
+
+#undef DEB_PREFIX
+#define DEB_PREFIX "uver"
+
+#include <asm/current.h>
+
+#include "ehca_classes.h"
+#include "ehca_iverbs.h"
+#include "ehca_kernel.h"
+#include "ehca_mrmw.h"
+#include "ehca_tools.h"
+#include "hcp_if.h"
+
+struct ib_ucontext *ehca_alloc_ucontext(struct ib_device *device,
+					struct ib_udata *udata)
+{
+	struct ehca_ucontext *my_context = NULL;
+
+	EHCA_CHECK_ADR_P(device);
+	EDEB_EN(7, "device=%p name=%s", device, device->name);
+
+	my_context = kzalloc(sizeof *my_context, GFP_KERNEL);
+	if (NULL == my_context) {
+		EDEB_ERR(4, "Out of memory device=%p", device);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	EDEB_EX(7, "device=%p ucontext=%p", device, my_context);
+
+	return &my_context->ib_ucontext;
+}
+
+int ehca_dealloc_ucontext(struct ib_ucontext *context)
+{
+	struct ehca_ucontext *my_context = NULL;
+	EHCA_CHECK_ADR(context);
+	EDEB_EN(7, "ucontext=%p", context);
+	my_context = container_of(context, struct ehca_ucontext, ib_ucontext);
+	kfree(my_context);
+	EDEB_EN(7, "ucontext=%p", context);
+	return 0;
+}
+
+struct page *ehca_nopage(struct vm_area_struct *vma,
+			 unsigned long address, int *type)
+{
+	struct page *mypage = NULL;
+	u64 fileoffset = vma->vm_pgoff << PAGE_SHIFT;
+	u32 idr_handle = fileoffset >> 32;
+	u32 q_type = (fileoffset >> 28) & 0xF;	  /* CQ, QP,...        */
+	u32 rsrc_type = (fileoffset >> 24) & 0xF; /* sq,rq,cmnd_window */
+	u32 cur_pid = current->tgid;
+	unsigned long flags;
+
+	EDEB_EN(7,
+		"vm_start=%lx vm_end=%lx vm_page_prot=%lx vm_fileoff=%lx "
+		"address=%lx",
+		vma->vm_start, vma->vm_end, vma->vm_page_prot, fileoffset,
+		address);
+
+	if (q_type == 1) { /* CQ */
+		struct ehca_cq *cq = NULL;
+		u64 offset;
+		void *vaddr = NULL;
+
+		spin_lock_irqsave(&ehca_cq_idr_lock, flags);
+		cq = idr_find(&ehca_cq_idr, idr_handle);
+		spin_unlock_irqrestore(&ehca_cq_idr_lock, flags);
+
+                if (cq->ownpid!=cur_pid) {
+                        EDEB_ERR(4, "Invalid caller pid=%x ownpid=%x",
+                                 cur_pid, cq->ownpid);
+                        return NOPAGE_SIGBUS;
+                }
+
+		/* make sure this mmap really belongs to the authorized user */
+		if (cq == 0) {
+			EDEB_ERR(4, "cq is NULL ret=NOPAGE_SIGBUS");
+			return NOPAGE_SIGBUS;
+		}
+		if (rsrc_type == 2) {
+			EDEB(6, "cq=%p cq queuearea", cq);
+			offset = address - vma->vm_start;
+			vaddr = ipz_qeit_calc(&cq->ipz_queue, offset);
+			EDEB(6, "offset=%lx vaddr=%p", offset, vaddr);
+			mypage = virt_to_page(vaddr);
+		}
+	} else if (q_type == 2) { /* QP */
+                struct ehca_qp *qp = NULL;
+                struct ehca_pd *pd = NULL;
+		u64 offset;
+		void *vaddr = NULL;
+
+		spin_lock_irqsave(&ehca_qp_idr_lock, flags);
+		qp = idr_find(&ehca_qp_idr, idr_handle);
+		spin_unlock_irqrestore(&ehca_qp_idr_lock, flags);
+
+
+                pd = container_of(qp->ib_qp.pd, struct ehca_pd, ib_pd);
+                if (pd->ownpid!=cur_pid) {
+                        EDEB_ERR(4, "Invalid caller pid=%x ownpid=%x",
+                                 cur_pid, pd->ownpid);
+                        return NOPAGE_SIGBUS;
+                }
+
+		/* make sure this mmap really belongs to the authorized user */
+		if (qp == NULL) {
+			EDEB_ERR(4, "qp is NULL ret=NOPAGE_SIGBUS");
+			return NOPAGE_SIGBUS;
+		}
+		if (rsrc_type == 2) {	/* rqueue */
+			EDEB(6, "qp=%p qp rqueuearea", qp);
+			offset = address - vma->vm_start;
+			vaddr = ipz_qeit_calc(&qp->ipz_rqueue, offset);
+			EDEB(6, "offset=%lx vaddr=%p", offset, vaddr);
+			mypage = virt_to_page(vaddr);
+		} else if (rsrc_type == 3) {	/* squeue */
+			EDEB(6, "qp=%p qp squeuearea", qp);
+			offset = address - vma->vm_start;
+			vaddr = ipz_qeit_calc(&qp->ipz_squeue, offset);
+			EDEB(6, "offset=%lx vaddr=%p", offset, vaddr);
+			mypage = virt_to_page(vaddr);
+		}
+	}
+
+	if (mypage == NULL) {
+		EDEB_ERR(4, "Invalid page adr==NULL ret=NOPAGE_SIGBUS");
+		return NOPAGE_SIGBUS;
+	}
+	get_page(mypage);
+	EDEB_EX(7, "page adr=%p", mypage);
+	return mypage;
+}
+
+static struct vm_operations_struct ehcau_vm_ops = {
+	.nopage = ehca_nopage,
+};
+
+int ehca_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
+{
+	u64 fileoffset = vma->vm_pgoff << PAGE_SHIFT;
+
+
+	u32 idr_handle = fileoffset >> 32;
+	u32 q_type = (fileoffset >> 28) & 0xF;	  /* CQ, QP,...        */
+	u32 rsrc_type = (fileoffset >> 24) & 0xF; /* sq,rq,cmnd_window */
+	u32 ret = -EFAULT;	/* assume the worst             */
+	u64 vsize = 0;		/* must be calculated/set below */
+	u64 physical = 0;	/* must be calculated/set below */
+        u32 cur_pid = current->tgid;
+	unsigned long flags;
+
+	EDEB_EN(7, "vm_start=%lx vm_end=%lx vm_page_prot=%lx vm_fileoff=%lx",
+		vma->vm_start, vma->vm_end, vma->vm_page_prot, fileoffset);
+
+	if (q_type == 1) { /* CQ */
+		struct ehca_cq *cq;
+
+		spin_lock_irqsave(&ehca_cq_idr_lock, flags);
+		cq = idr_find(&ehca_cq_idr, idr_handle);
+		spin_unlock_irqrestore(&ehca_cq_idr_lock, flags);
+
+                if (cq->ownpid!=cur_pid) {
+                        EDEB_ERR(4, "Invalid caller pid=%x ownpid=%x",
+                                 cur_pid, cq->ownpid);
+                        return -ENOMEM;
+                }
+
+		/* make sure this mmap really belongs to the authorized user */
+		if (cq == 0)
+			return -EINVAL;
+		if (cq->ib_cq.uobject == 0)
+			return -EINVAL;
+		if (cq->ib_cq.uobject->context != context)
+			return -EINVAL;
+		if (rsrc_type == 1) {	/* galpa fw handle */
+			EDEB(6, "cq=%p cq triggerarea", cq);
+			vma->vm_flags |= VM_RESERVED;
+			vsize = vma->vm_end - vma->vm_start;
+			if (vsize != 4096) {
+				EDEB_ERR(4, "invalid vsize=%lx",
+					 vma->vm_end - vma->vm_start);
+				ret = -EINVAL;
+				goto mmap_exit0;
+			}
+
+			physical = cq->galpas.user.fw_handle;
+			vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+			vma->vm_flags |= VM_IO | VM_RESERVED;
+
+			EDEB(6, "vsize=%lx physical=%lx", vsize,
+			     physical);
+			ret =
+			    remap_pfn_range(vma, vma->vm_start,
+					    physical >> PAGE_SHIFT, vsize,
+					    vma->vm_page_prot);
+			if (ret != 0) {
+				EDEB_ERR(4,
+					 "Error: remap_pfn_range() returned %x!",
+					 ret);
+				ret = -ENOMEM;
+			}
+			goto mmap_exit0;
+		} else if (rsrc_type == 2) {	/* cq queue_addr */
+			EDEB(6, "cq=%p cq q_addr", cq);
+			/* vma->vm_page_prot =
+			 * pgprot_noncached(vma->vm_page_prot); */
+			vma->vm_flags |= VM_RESERVED;
+			vma->vm_ops = &ehcau_vm_ops;
+			ret = 0;
+			goto mmap_exit0;
+		} else {
+			EDEB_ERR(6, "bad resource type %x", rsrc_type);
+			ret = -EINVAL;
+			goto mmap_exit0;
+		}
+	} else if (q_type == 2) { /* QP */
+                struct ehca_qp *qp = NULL;
+                struct ehca_pd *pd = NULL;
+
+		spin_lock_irqsave(&ehca_qp_idr_lock, flags);
+		qp = idr_find(&ehca_qp_idr, idr_handle);
+		spin_unlock_irqrestore(&ehca_qp_idr_lock, flags);
+
+
+                pd = container_of(qp->ib_qp.pd, struct ehca_pd, ib_pd);
+                if (pd->ownpid!=cur_pid) {
+                        EDEB_ERR(4, "Invalid caller pid=%x ownpid=%x",
+                                 cur_pid, pd->ownpid);
+                        return -ENOMEM;
+                }
+
+		/* make sure this mmap really belongs to the authorized user */
+		if (qp == NULL || qp->ib_qp.uobject == NULL ||
+		    qp->ib_qp.uobject->context != context) {
+			EDEB(6, "qp=%p, uobject=%p, context=%p",
+			     qp, qp->ib_qp.uobject, qp->ib_qp.uobject->context);
+			ret = -EINVAL;
+			goto mmap_exit0;
+		}
+		if (rsrc_type == 1) {	/* galpa fw handle */
+			EDEB(6, "qp=%p qp triggerarea", qp);
+			vma->vm_flags |= VM_RESERVED;
+			vsize = vma->vm_end - vma->vm_start;
+			if (vsize != 4096) {
+				EDEB_ERR(4, "invalid vsize=%lx",
+					 vma->vm_end - vma->vm_start);
+				ret = -EINVAL;
+				goto mmap_exit0;
+			}
+
+			physical = qp->galpas.user.fw_handle;
+			vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+			vma->vm_flags |= VM_IO | VM_RESERVED;
+
+			EDEB(6, "vsize=%lx physical=%lx", vsize,
+			     physical);
+			ret =
+			    remap_pfn_range(vma, vma->vm_start,
+					    physical >> PAGE_SHIFT, vsize,
+					    vma->vm_page_prot);
+			if (ret != 0) {
+				EDEB_ERR(4,
+					 "Error: remap_pfn_range() returned %x!",
+					 ret);
+				ret = -ENOMEM;
+			}
+			goto mmap_exit0;
+		} else if (rsrc_type == 2) {	/* qp rqueue_addr */
+			EDEB(6, "qp=%p qp rqueue_addr", qp);
+			vma->vm_flags |= VM_RESERVED;
+			vma->vm_ops = &ehcau_vm_ops;
+			ret = 0;
+			goto mmap_exit0;
+		} else if (rsrc_type == 3) {	/* qp squeue_addr */
+			EDEB(6, "qp=%p qp squeue_addr", qp);
+			vma->vm_flags |= VM_RESERVED;
+			vma->vm_ops = &ehcau_vm_ops;
+			ret = 0;
+			goto mmap_exit0;
+		} else {
+			EDEB_ERR(4, "bad resource type %x",
+				 rsrc_type);
+			ret = -EINVAL;
+			goto mmap_exit0;
+		}
+	} else {
+		EDEB_ERR(4, "bad queue type %x", q_type);
+		ret = -EINVAL;
+		goto mmap_exit0;
+	}
+
+mmap_exit0:
+	EDEB_EX(7, "ret=%x", ret);
+	return ret;
+}
+
+int ehca_mmap_nopage(u64 foffset,u64 length,void ** mapped,struct vm_area_struct ** vma)
+{
+	EDEB_EN(7, "foffset=%lx length=%lx", foffset, length);
+	down_write(&current->mm->mmap_sem);
+	*mapped=(void*)
+		do_mmap(NULL,0,
+			length,
+			PROT_WRITE, MAP_SHARED|MAP_ANONYMOUS,
+			foffset);
+	up_write(&current->mm->mmap_sem);
+	if (*mapped) {
+		*vma = find_vma(current->mm,(u64)*mapped);
+		if (*vma) {
+			(*vma)->vm_flags |= VM_RESERVED;
+			(*vma)->vm_ops = &ehcau_vm_ops;
+		} else {
+			EDEB_ERR(4,"couldn't find queue vma queue=%p",
+				 *mapped);
+		}
+	} else {
+		EDEB_ERR(4,"couldn't create mmap length=%lx",length);
+	}
+	EDEB_EX(7, "mapped=%p", *mapped);
+	return 0;
+}
+
+int ehca_mmap_register(u64 physical,void ** mapped,struct vm_area_struct ** vma)
+{
+	int ret;
+	unsigned long vsize;
+	ehca_mmap_nopage(0,4096,mapped,vma);
+	(*vma)->vm_flags |= VM_RESERVED;
+	vsize = (*vma)->vm_end - (*vma)->vm_start;
+	if (vsize != 4096) {
+		EDEB_ERR(4, "invalid vsize=%lx",
+			 (*vma)->vm_end - (*vma)->vm_start);
+		ret = -EINVAL;
+		return ret;
+	}
+
+	(*vma)->vm_page_prot = pgprot_noncached((*vma)->vm_page_prot);
+	(*vma)->vm_flags |= VM_IO | VM_RESERVED;
+
+	EDEB(6, "vsize=%lx physical=%lx", vsize,
+	     physical);
+	ret =
+		remap_pfn_range((*vma), (*vma)->vm_start,
+				physical >> PAGE_SHIFT, vsize,
+				(*vma)->vm_page_prot);
+	if (ret != 0) {
+		EDEB_ERR(4,
+			 "Error: remap_pfn_range() returned %x!",
+			 ret);
+		ret = -ENOMEM;
+	}
+	return ret;
+
+}
+
+int ehca_munmap(unsigned long addr, size_t len) {
+	int ret=0;
+	struct mm_struct *mm = current->mm;
+	if (mm!=0) {
+		down_write(&mm->mmap_sem);
+		ret = do_munmap(mm, addr, len);
+		up_write(&mm->mmap_sem);
+	}
+	return ret;
+}

Re: [PATCH 04/16] ehca: userspace support

[Jörn Engel]

More minors.

On Thu, 27 April 2006 12:48:22 +0200, Heiko J Schick wrote:
> +
> +	EDEB_EN(7,
> +		"vm_start=%lx vm_end=%lx vm_page_prot=%lx vm_fileoff=%lx "
> +		"address=%lx",
> +		vma->vm_start, vma->vm_end, vma->vm_page_prot, fileoffset,
> +		address);

Gesundheit!  Seriously, I suspect "EDEB_EN" is not the best possible
name to pick.

> +                if (cq->ownpid!=cur_pid) {

Coding style would require spaces around binary operators.

Jörn

-- 
He that composes himself is wiser than he that composes a book.
-- B. Franklin

Re: [PATCH 04/16] ehca: userspace support

[Michael Ellerman]

[-- Attachment #1: Type: text/plain, Size: 709 bytes --]

On Thu, 2006-04-27 at 13:43 +0200, Jörn Engel wrote:
> More minors.
> 
> On Thu, 27 April 2006 12:48:22 +0200, Heiko J Schick wrote:
> > +
> > +	EDEB_EN(7,
> > +		"vm_start=%lx vm_end=%lx vm_page_prot=%lx vm_fileoff=%lx "
> > +		"address=%lx",
> > +		vma->vm_start, vma->vm_end, vma->vm_page_prot, fileoffset,
> > +		address);
> 
> Gesundheit!  Seriously, I suspect "EDEB_EN" is not the best possible
> name to pick.

Try pr_debug() in include/linux/kernel.h

cheers

-- 
Michael Ellerman
IBM OzLabs

wwweb: http://michael.ellerman.id.au
phone: +61 2 6212 1183 (tie line 70 21183)

We do not inherit the earth from our ancestors,
we borrow it from our children. - S.M.A.R.T Person

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 191 bytes --]

Re: [openib-general] Re: [PATCH 04/16] ehca: userspace support

[Heiko J Schick]

Hello Michael,

On 28.04.2006, at 00:36, Michael Ellerman wrote:

> Try pr_debug() in include/linux/kernel.h

The problem I see with pr_debug() is that it could only activated via
a compile flag. To use the debug outputs you have to re-compile /  
compile
your own kernel.

Regards,
	Heiko

Re: [openib-general] Re: [PATCH 04/16] ehca: userspace support

[Pekka Enberg]

Hi Heiko,

On 4/28/06, Heiko J Schick <info@schihei.de> wrote:
> The problem I see with pr_debug() is that it could only activated via
> a compile flag. To use the debug outputs you have to re-compile /
> compile your own kernel.

Do you really need this heavy debug logging in the first place? You
can use kprobes for arbitrary run-time inspection anyway, so logging
everything seems wasteful.

                                             Pekka

Re: [openib-general] Re: [PATCH 04/16] ehca: userspace support

[Michael Ellerman]

[-- Attachment #1: Type: text/plain, Size: 924 bytes --]

On Fri, 2006-04-28 at 09:32 +0300, Pekka Enberg wrote:
> Hi Heiko,
> 
> On 4/28/06, Heiko J Schick <info@schihei.de> wrote:
> > The problem I see with pr_debug() is that it could only activated via
> > a compile flag. To use the debug outputs you have to re-compile /
> > compile your own kernel.
> 
> Do you really need this heavy debug logging in the first place? You
> can use kprobes for arbitrary run-time inspection anyway, so logging
> everything seems wasteful.

Yeah, I really don't think you want to be running with that kind of
debugging gunk in a production kernel. Is someone who can't build their
own kernel really going to be able to make sense of the output?

cheers

-- 
Michael Ellerman
IBM OzLabs

wwweb: http://michael.ellerman.id.au
phone: +61 2 6212 1183 (tie line 70 21183)

We do not inherit the earth from our ancestors,
we borrow it from our children. - S.M.A.R.T Person

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 191 bytes --]

Re: [openib-general] Re: [PATCH 04/16] ehca: userspace support

[Heiko J Schick]

Hello,

On 28.04.2006, at 08:32, Pekka Enberg wrote:

>> The problem I see with pr_debug() is that it could only activated via
>> a compile flag. To use the debug outputs you have to re-compile /
>> compile your own kernel.
>
> Do you really need this heavy debug logging in the first place? You
> can use kprobes for arbitrary run-time inspection anyway, so logging
> everything seems wasteful.

The problem I see with kprobes is that you have to set several kernel
configuration options (e.g. CONFIG_KPROBES, CONFIG_DEBUG_INFO, etc.)
on compile time to use it. Same problem with pr_debug().

Regards,
	Heiko

Re: [openib-general] Re: [PATCH 04/16] ehca: userspace support

[Heiko Carstens]

> >>The problem I see with pr_debug() is that it could only activated via
> >>a compile flag. To use the debug outputs you have to re-compile /
> >>compile your own kernel.
> >
> >Do you really need this heavy debug logging in the first place? You
> >can use kprobes for arbitrary run-time inspection anyway, so logging
> >everything seems wasteful.
> 
> The problem I see with kprobes is that you have to set several kernel
> configuration options (e.g. CONFIG_KPROBES, CONFIG_DEBUG_INFO, etc.)
> on compile time to use it. Same problem with pr_debug().

It might be worth to move the s390 debug feature to common code. At least
it has proven many times to be very useful in device driver debugging...
See Documentation/s390/s390dbf.txt and arch/s390/kernel/debug.c.