Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org A07C822DC1
From:   Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To:     linux-kernel@vger.kernel.org
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        stable@vger.kernel.org, Eloy Coto Pereiro <eloy.coto@gmail.com>,
        Hans de Goede <hdegoede@redhat.com>
Subject: [PATCH 4.16 092/113] virt: vbox: Use __get_free_pages instead of kmalloc for DMA32 memory
Date:   Mon, 30 Apr 2018 12:25:03 -0700
Message-Id: <20180430184019.089903150@linuxfoundation.org>
In-Reply-To: <20180430184015.043892819@linuxfoundation.org>
References: <20180430184015.043892819@linuxfoundation.org>
User-Agent: quilt/0.65
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

4.16-stable review patch.  If anyone has any objections, please let me know.

------------------

From: Hans de Goede <hdegoede@redhat.com>

commit faf6a2a44164c0fb2c2a82692ab9051917514bce upstream.

It is not possible to get DMA32 zone memory through kmalloc, causing
the vboxguest driver to malfunction due to getting memory above
4G which the PCI device cannot handle.

This commit changes the kmalloc calls where the 4G limit matters to
using __get_free_pages() fixing vboxguest not working on x86_64 guests
with more then 4G RAM.

Cc: stable@vger.kernel.org
Reported-by: Eloy Coto Pereiro <eloy.coto@gmail.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

---
 drivers/virt/vboxguest/vboxguest_linux.c |   19 ++++++++++++++++---
 drivers/virt/vboxguest/vboxguest_utils.c |    5 +++--
 2 files changed, 19 insertions(+), 5 deletions(-)

--- a/drivers/virt/vboxguest/vboxguest_linux.c
+++ b/drivers/virt/vboxguest/vboxguest_linux.c
@@ -87,6 +87,7 @@ static long vbg_misc_device_ioctl(struct
 	struct vbg_session *session = filp->private_data;
 	size_t returned_size, size;
 	struct vbg_ioctl_hdr hdr;
+	bool is_vmmdev_req;
 	int ret = 0;
 	void *buf;
 
@@ -106,8 +107,17 @@ static long vbg_misc_device_ioctl(struct
 	if (size > SZ_16M)
 		return -E2BIG;
 
-	/* __GFP_DMA32 because IOCTL_VMMDEV_REQUEST passes this to the host */
-	buf = kmalloc(size, GFP_KERNEL | __GFP_DMA32);
+	/*
+	 * IOCTL_VMMDEV_REQUEST needs the buffer to be below 4G to avoid
+	 * the need for a bounce-buffer and another copy later on.
+	 */
+	is_vmmdev_req = (req & ~IOCSIZE_MASK) == VBG_IOCTL_VMMDEV_REQUEST(0) ||
+			 req == VBG_IOCTL_VMMDEV_REQUEST_BIG;
+
+	if (is_vmmdev_req)
+		buf = vbg_req_alloc(size, VBG_IOCTL_HDR_TYPE_DEFAULT);
+	else
+		buf = kmalloc(size, GFP_KERNEL);
 	if (!buf)
 		return -ENOMEM;
 
@@ -132,7 +142,10 @@ static long vbg_misc_device_ioctl(struct
 		ret = -EFAULT;
 
 out:
-	kfree(buf);
+	if (is_vmmdev_req)
+		vbg_req_free(buf, size);
+	else
+		kfree(buf);
 
 	return ret;
 }
--- a/drivers/virt/vboxguest/vboxguest_utils.c
+++ b/drivers/virt/vboxguest/vboxguest_utils.c
@@ -65,8 +65,9 @@ VBG_LOG(vbg_debug, pr_debug);
 void *vbg_req_alloc(size_t len, enum vmmdev_request_type req_type)
 {
 	struct vmmdev_request_header *req;
+	int order = get_order(PAGE_ALIGN(len));
 
-	req = kmalloc(len, GFP_KERNEL | __GFP_DMA32);
+	req = (void *)__get_free_pages(GFP_KERNEL | GFP_DMA32, order);
 	if (!req)
 		return NULL;
 
@@ -87,7 +88,7 @@ void vbg_req_free(void *req, size_t len)
 	if (!req)
 		return;
 
-	kfree(req);
+	free_pages((unsigned long)req, get_order(PAGE_ALIGN(len)));
 }
 
 /* Note this function returns a VBox status code, not a negative errno!! */