Merge ../from-linus

Documentation/usb/usbmon.txt

@@ -102,7 +102,7 @@ Here is the list of words, from left to right:
 - URB Status. This field makes no sense for submissions, but is present
   to help scripts with parsing. In error case, it contains the error code.
   In case of a setup packet, it contains a Setup Tag. If scripts read a number
-  in this field, the proceed to read Data Length. Otherwise, they read
+  in this field, they proceed to read Data Length. Otherwise, they read
   the setup packet before reading the Data Length.
 - Setup packet, if present, consists of 5 words: one of each for bmRequestType,
   bRequest, wValue, wIndex, wLength, as specified by the USB Specification 2.0.

arch/alpha/kernel/pci.c

@@ -350,8 +350,24 @@ pcibios_resource_to_bus(struct pci_dev *dev, struct pci_bus_region *region,
 	region->end = res->end - offset;
 }
 
+void pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+			     struct pci_bus_region *region)
+{
+	struct pci_controller *hose = (struct pci_controller *)dev->sysdata;
+	unsigned long offset = 0;
+
+	if (res->flags & IORESOURCE_IO)
+		offset = hose->io_space->start;
+	else if (res->flags & IORESOURCE_MEM)
+		offset = hose->mem_space->start;
+
+	res->start = region->start + offset;
+	res->end = region->end + offset;
+}
+
 #ifdef CONFIG_HOTPLUG
 EXPORT_SYMBOL(pcibios_resource_to_bus);
+EXPORT_SYMBOL(pcibios_bus_to_resource);
 #endif
 
 int

arch/arm/kernel/bios32.c

@@ -447,9 +447,26 @@ pcibios_resource_to_bus(struct pci_dev *dev, struct pci_bus_region *region,
 	region->end   = res->end - offset;
 }
 
+void __devinit
+pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+			struct pci_bus_region *region)
+{
+	struct pci_sys_data *root = dev->sysdata;
+	unsigned long offset = 0;
+
+	if (res->flags & IORESOURCE_IO)
+		offset = root->io_offset;
+	if (res->flags & IORESOURCE_MEM)
+		offset = root->mem_offset;
+
+	res->start = region->start + offset;
+	res->end   = region->end + offset;
+}
+
 #ifdef CONFIG_HOTPLUG
 EXPORT_SYMBOL(pcibios_fixup_bus);
 EXPORT_SYMBOL(pcibios_resource_to_bus);
+EXPORT_SYMBOL(pcibios_bus_to_resource);
 #endif
 
 /*

arch/ppc/kernel/pci.c

@@ -160,6 +160,21 @@ void pcibios_resource_to_bus(struct pci_dev *dev, struct pci_bus_region *region,
 }
 EXPORT_SYMBOL(pcibios_resource_to_bus);
 
+void pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+			     struct pci_bus_region *region)
+{
+	unsigned long offset = 0;
+	struct pci_controller *hose = dev->sysdata;
+
+	if (hose && res->flags & IORESOURCE_IO)
+		offset = (unsigned long)hose->io_base_virt - isa_io_base;
+	else if (hose && res->flags & IORESOURCE_MEM)
+		offset = hose->pci_mem_offset;
+	res->start = region->start + offset;
+	res->end = region->end + offset;
+}
+EXPORT_SYMBOL(pcibios_bus_to_resource);
+
 /*
  * We need to avoid collisions with `mirrored' VGA ports
  * and other strange ISA hardware, so we always want the

arch/ppc64/kernel/pci.c

@@ -108,8 +108,28 @@ void  pcibios_resource_to_bus(struct pci_dev *dev, struct pci_bus_region *region
 	region->end = res->end - offset;
 }
 
+void pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+			      struct pci_bus_region *region)
+{
+	unsigned long offset = 0;
+	struct pci_controller *hose = pci_bus_to_host(dev->bus);
+
+	if (!hose)
+		return;
+
+	if (res->flags & IORESOURCE_IO)
+	        offset = (unsigned long)hose->io_base_virt - pci_io_base;
+
+	if (res->flags & IORESOURCE_MEM)
+		offset = hose->pci_mem_offset;
+
+	res->start = region->start + offset;
+	res->end = region->end + offset;
+}
+
 #ifdef CONFIG_HOTPLUG
 EXPORT_SYMBOL(pcibios_resource_to_bus);
+EXPORT_SYMBOL(pcibios_bus_to_resource);
 #endif
 
 /*

arch/sparc64/kernel/pci.c

@@ -413,6 +413,12 @@ static int pci_assign_bus_resource(const struct pci_bus *bus,
 	return -EBUSY;
 }
 
+void pci_update_resource(struct pci_dev *dev, struct resource *res, int resno)
+{
+	/* Not implemented for sparc64... */
+	BUG();
+}
+
 int pci_assign_resource(struct pci_dev *pdev, int resource)
 {
 	struct pcidev_cookie *pcp = pdev->sysdata;

drivers/char/rtc.c

@@ -1209,6 +1209,7 @@ static int rtc_proc_open(struct inode *inode, struct file *file)
 
 void rtc_get_rtc_time(struct rtc_time *rtc_tm)
 {
+	unsigned long uip_watchdog = jiffies;
 	unsigned char ctrl;
 #ifdef CONFIG_MACH_DECSTATION
 	unsigned int real_year;
@@ -1224,8 +1225,10 @@ void rtc_get_rtc_time(struct rtc_time *rtc_tm)
 	 * Once the read clears, read the RTC time (again via ioctl). Easy.
 	 */
 
-	if (rtc_is_updating() != 0)
-		msleep(20);
+	while (rtc_is_updating() != 0 && jiffies - uip_watchdog < 2*HZ/100) {
+		barrier();
+		cpu_relax();
+	}
 
 	/*
 	 * Only the values that we read from the RTC are set. We leave

drivers/pci/pci.c

@@ -221,6 +221,37 @@ pci_find_parent_resource(const struct pci_dev *dev, struct resource *res)
 	return best;
 }
 
+/**
+ * pci_restore_bars - restore a devices BAR values (e.g. after wake-up)
+ * @dev: PCI device to have its BARs restored
+ *
+ * Restore the BAR values for a given device, so as to make it
+ * accessible by its driver.
+ */
+void
+pci_restore_bars(struct pci_dev *dev)
+{
+	int i, numres;
+
+	switch (dev->hdr_type) {
+	case PCI_HEADER_TYPE_NORMAL:
+		numres = 6;
+		break;
+	case PCI_HEADER_TYPE_BRIDGE:
+		numres = 2;
+		break;
+	case PCI_HEADER_TYPE_CARDBUS:
+		numres = 1;
+		break;
+	default:
+		/* Should never get here, but just in case... */
+		return;
+	}
+
+	for (i = 0; i < numres; i ++)
+		pci_update_resource(dev, &dev->resource[i], i);
+}
+
 /**
  * pci_set_power_state - Set the power state of a PCI device
  * @dev: PCI device to be suspended
@@ -239,7 +270,7 @@ int (*platform_pci_set_power_state)(struct pci_dev *dev, pci_power_t t);
 int
 pci_set_power_state(struct pci_dev *dev, pci_power_t state)
 {
-	int pm;
+	int pm, need_restore = 0;
 	u16 pmcsr, pmc;
 
 	/* bound the state we're entering */
@@ -278,14 +309,17 @@ pci_set_power_state(struct pci_dev *dev, pci_power_t state)
 			return -EIO;
 	}
 
+	pci_read_config_word(dev, pm + PCI_PM_CTRL, &pmcsr);
+
 	/* If we're in D3, force entire word to 0.
 	 * This doesn't affect PME_Status, disables PME_En, and
 	 * sets PowerState to 0.
 	 */
-	if (dev->current_state >= PCI_D3hot)
+	if (dev->current_state >= PCI_D3hot) {
+		if (!(pmcsr & PCI_PM_CTRL_NO_SOFT_RESET))
+			need_restore = 1;
 		pmcsr = 0;
-	else {
-		pci_read_config_word(dev, pm + PCI_PM_CTRL, &pmcsr);
+	} else {
 		pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
 		pmcsr |= state;
 	}
@@ -308,6 +342,22 @@ pci_set_power_state(struct pci_dev *dev, pci_power_t state)
 		platform_pci_set_power_state(dev, state);
 
 	dev->current_state = state;
+
+	/* According to section 5.4.1 of the "PCI BUS POWER MANAGEMENT
+	 * INTERFACE SPECIFICATION, REV. 1.2", a device transitioning
+	 * from D3hot to D0 _may_ perform an internal reset, thereby
+	 * going to "D0 Uninitialized" rather than "D0 Initialized".
+	 * For example, at least some versions of the 3c905B and the
+	 * 3c556B exhibit this behaviour.
+	 *
+	 * At least some laptop BIOSen (e.g. the Thinkpad T21) leave
+	 * devices in a D3hot state at boot.  Consequently, we need to
+	 * restore at least the BARs so that the device will be
+	 * accessible to its driver.
+	 */
+	if (need_restore)
+		pci_restore_bars(dev);
+
 	return 0;
 }
 
@@ -805,6 +855,7 @@ struct pci_dev *isa_bridge;
 EXPORT_SYMBOL(isa_bridge);
 #endif
 
+EXPORT_SYMBOL_GPL(pci_restore_bars);
 EXPORT_SYMBOL(pci_enable_device_bars);
 EXPORT_SYMBOL(pci_enable_device);
 EXPORT_SYMBOL(pci_disable_device);

drivers/pci/setup-res.c

@@ -26,7 +26,7 @@
 #include "pci.h"
 
 
-static void
+void
 pci_update_resource(struct pci_dev *dev, struct resource *res, int resno)
 {
 	struct pci_bus_region region;

drivers/pcmcia/yenta_socket.c

@@ -605,9 +605,8 @@ static int yenta_search_res(struct yenta_socket *socket, struct resource *res,
 
 static void yenta_allocate_res(struct yenta_socket *socket, int nr, unsigned type, int addr_start, int addr_end)
 {
-	struct pci_bus *bus;
 	struct resource *root, *res;
-	u32 start, end;
+	struct pci_bus_region region;
 	unsigned mask;
 
 	res = socket->dev->resource + PCI_BRIDGE_RESOURCES + nr;
@@ -620,15 +619,13 @@ static void yenta_allocate_res(struct yenta_socket *socket, int nr, unsigned typ
 	if (type & IORESOURCE_IO)
 		mask = ~3;
 
-	bus = socket->dev->subordinate;
-	res->name = bus->name;
+	res->name = socket->dev->subordinate->name;
 	res->flags = type;
 
-	start = config_readl(socket, addr_start) & mask;
-	end = config_readl(socket, addr_end) | ~mask;
-	if (start && end > start && !override_bios) {
-		res->start = start;
-		res->end = end;
+	region.start = config_readl(socket, addr_start) & mask;
+	region.end = config_readl(socket, addr_end) | ~mask;
+	if (region.start && region.end > region.start && !override_bios) {
+		pcibios_bus_to_resource(socket->dev, res, &region);
 		root = pci_find_parent_resource(socket->dev, res);
 		if (root && (request_resource(root, res) == 0))
 			return;

drivers/usb/host/ehci-dbg.c

@@ -527,7 +527,7 @@ show_periodic (struct class_device *class_dev, char *buf)
 						p.qh->period,
 						le32_to_cpup (&p.qh->hw_info2)
 							/* uframe masks */
-							& 0xffff,
+							& (QH_CMASK | QH_SMASK),
 						p.qh);
 				size -= temp;
 				next += temp;

drivers/usb/host/ehci-q.c

@@ -222,7 +222,7 @@ __acquires(ehci->lock)
 		struct ehci_qh	*qh = (struct ehci_qh *) urb->hcpriv;
 
 		/* S-mask in a QH means it's an interrupt urb */
-		if ((qh->hw_info2 & __constant_cpu_to_le32 (0x00ff)) != 0) {
+		if ((qh->hw_info2 & __constant_cpu_to_le32 (QH_SMASK)) != 0) {
 
 			/* ... update hc-wide periodic stats (for usbfs) */
 			ehci_to_hcd(ehci)->self.bandwidth_int_reqs--;
@@ -428,7 +428,8 @@ qh_completions (struct ehci_hcd *ehci, struct ehci_qh *qh, struct pt_regs *regs)
 			/* should be rare for periodic transfers,
 			 * except maybe high bandwidth ...
 			 */
-			if (qh->period) {
+			if ((__constant_cpu_to_le32 (QH_SMASK)
+					& qh->hw_info2) != 0) {
 				intr_deschedule (ehci, qh);
 				(void) qh_schedule (ehci, qh);
 			} else

drivers/usb/host/ehci-sched.c

@@ -301,7 +301,7 @@ static int qh_link_periodic (struct ehci_hcd *ehci, struct ehci_qh *qh)
 
 	dev_dbg (&qh->dev->dev,
 		"link qh%d-%04x/%p start %d [%d/%d us]\n",
-		period, le32_to_cpup (&qh->hw_info2) & 0xffff,
+		period, le32_to_cpup (&qh->hw_info2) & (QH_CMASK | QH_SMASK),
 		qh, qh->start, qh->usecs, qh->c_usecs);
 
 	/* high bandwidth, or otherwise every microframe */
@@ -385,7 +385,8 @@ static void qh_unlink_periodic (struct ehci_hcd *ehci, struct ehci_qh *qh)
 
 	dev_dbg (&qh->dev->dev,
 		"unlink qh%d-%04x/%p start %d [%d/%d us]\n",
-		qh->period, le32_to_cpup (&qh->hw_info2) & 0xffff,
+		qh->period,
+		le32_to_cpup (&qh->hw_info2) & (QH_CMASK | QH_SMASK),
 		qh, qh->start, qh->usecs, qh->c_usecs);
 
 	/* qh->qh_next still "live" to HC */
@@ -411,7 +412,7 @@ static void intr_deschedule (struct ehci_hcd *ehci, struct ehci_qh *qh)
 	 * active high speed queues may need bigger delays...
 	 */
 	if (list_empty (&qh->qtd_list)
-			|| (__constant_cpu_to_le32 (0x0ff << 8)
+			|| (__constant_cpu_to_le32 (QH_CMASK)
 					& qh->hw_info2) != 0)
 		wait = 2;
 	else
@@ -533,7 +534,7 @@ static int qh_schedule (struct ehci_hcd *ehci, struct ehci_qh *qh)
 
 	/* reuse the previous schedule slots, if we can */
 	if (frame < qh->period) {
-		uframe = ffs (le32_to_cpup (&qh->hw_info2) & 0x00ff);
+		uframe = ffs (le32_to_cpup (&qh->hw_info2) & QH_SMASK);
 		status = check_intr_schedule (ehci, frame, --uframe,
 				qh, &c_mask);
 	} else {
@@ -569,10 +570,10 @@ static int qh_schedule (struct ehci_hcd *ehci, struct ehci_qh *qh)
 		qh->start = frame;
 
 		/* reset S-frame and (maybe) C-frame masks */
-		qh->hw_info2 &= __constant_cpu_to_le32 (~0xffff);
+		qh->hw_info2 &= __constant_cpu_to_le32(~(QH_CMASK | QH_SMASK));
 		qh->hw_info2 |= qh->period
 			? cpu_to_le32 (1 << uframe)
-			: __constant_cpu_to_le32 (0xff);
+			: __constant_cpu_to_le32 (QH_SMASK);
 		qh->hw_info2 |= c_mask;
 	} else
 		ehci_dbg (ehci, "reused qh %p schedule\n", qh);

drivers/usb/host/ehci.h

@@ -385,6 +385,11 @@ struct ehci_qh {
 	__le32			hw_info1;        /* see EHCI 3.6.2 */
 #define	QH_HEAD		0x00008000
 	__le32			hw_info2;        /* see EHCI 3.6.2 */
+#define	QH_SMASK	0x000000ff
+#define	QH_CMASK	0x0000ff00
+#define	QH_HUBADDR	0x007f0000
+#define	QH_HUBPORT	0x3f800000
+#define	QH_MULT		0xc0000000
 	__le32			hw_current;	 /* qtd list - see EHCI 3.6.4 */
 	
 	/* qtd overlay (hardware parts of a struct ehci_qtd) */

drivers/usb/host/isp116x-hcd.c

@@ -229,9 +229,11 @@ static void preproc_atl_queue(struct isp116x *isp116x)
 	struct isp116x_ep *ep;
 	struct urb *urb;
 	struct ptd *ptd;
-	u16 toggle = 0, dir = PTD_DIR_SETUP, len;
+	u16 len;
 
 	for (ep = isp116x->atl_active; ep; ep = ep->active) {
+		u16 toggle = 0, dir = PTD_DIR_SETUP;
+
 		BUG_ON(list_empty(&ep->hep->urb_list));
 		urb = container_of(ep->hep->urb_list.next,
 				   struct urb, urb_list);

drivers/usb/mon/Kconfig

@@ -9,9 +9,8 @@ config USB_MON
 	help
 	  If you say Y here, a component which captures the USB traffic
 	  between peripheral-specific drivers and HC drivers will be built.
-	  The USB_MON is similar in spirit and may be compatible with Dave
-	  Harding's USBMon.
+	  For more information, see <file:Documentation/usb/usbmon.txt>.
 
-	  This is somewhat experimental at this time, but it should be safe,
-	  as long as you aren't using modular USB and try to remove this
-	  module.
+	  This is somewhat experimental at this time, but it should be safe.
+
+	  If unsure, say Y.

drivers/usb/mon/Makefile

@@ -4,4 +4,5 @@
 
 usbmon-objs	:= mon_main.o mon_stat.o mon_text.o
 
+# This does not use CONFIG_USB_MON because we want this to use a tristate.
 obj-$(CONFIG_USB)	+= usbmon.o

fs/namei.c

@@ -1874,14 +1874,9 @@ int vfs_unlink(struct inode *dir, struct dentry *dentry)
 
 	/* We don't d_delete() NFS sillyrenamed files--they still exist. */
 	if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) {
-#if defined(CONFIG_INOTIFY) || defined(CONFIG_DNOTIFY)
-		dget(dentry);
+		struct inode *inode = dentry->d_inode;
 		d_delete(dentry);
-		fsnotify_unlink(dentry, dir);
-		dput(dentry);
-#else
-		d_delete(dentry);
-#endif
+		fsnotify_unlink(dentry, inode, dir);
 	}
 
 	return error;

include/asm-alpha/pci.h

@@ -251,6 +251,9 @@ static inline int pci_get_legacy_ide_irq(struct pci_dev *dev, int channel)
 extern void pcibios_resource_to_bus(struct pci_dev *, struct pci_bus_region *,
 				    struct resource *);
 
+extern void pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+				    struct pci_bus_region *region);
+
 #define pci_domain_nr(bus) ((struct pci_controller *)(bus)->sysdata)->index
 
 static inline int pci_proc_domain(struct pci_bus *bus)

include/asm-arm/pci.h

@@ -60,6 +60,10 @@ extern void
 pcibios_resource_to_bus(struct pci_dev *dev, struct pci_bus_region *region,
 			 struct resource *res);
 
+extern void
+pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+			struct pci_bus_region *region);
+
 static inline void pcibios_add_platform_entries(struct pci_dev *dev)
 {
 }

include/asm-generic/pci.h

@@ -22,6 +22,14 @@ pcibios_resource_to_bus(struct pci_dev *dev, struct pci_bus_region *region,
 	region->end = res->end;
 }
 
+static inline void
+pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+			struct pci_bus_region *region)
+{
+	res->start = region->start;
+	res->end = region->end;
+}
+
 #define pcibios_scan_all_fns(a, b)	0
 
 #ifndef HAVE_ARCH_PCI_GET_LEGACY_IDE_IRQ

include/asm-parisc/pci.h

@@ -253,6 +253,10 @@ extern void
 pcibios_resource_to_bus(struct pci_dev *dev, struct pci_bus_region *region,
 			 struct resource *res);
 
+extern void
+pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+			struct pci_bus_region *region);
+
 static inline void pcibios_add_platform_entries(struct pci_dev *dev)
 {
 }

include/asm-ppc/pci.h

@@ -105,6 +105,10 @@ extern void
 pcibios_resource_to_bus(struct pci_dev *dev, struct pci_bus_region *region,
 			struct resource *res);
 
+extern void
+pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+			struct pci_bus_region *region);
+
 extern void pcibios_add_platform_entries(struct pci_dev *dev);
 
 struct file;

include/asm-ppc64/pci.h

@@ -134,6 +134,10 @@ extern void
 pcibios_resource_to_bus(struct pci_dev *dev, struct pci_bus_region *region,
 			struct resource *res);
 
+extern void
+pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
+			struct pci_bus_region *region);
+
 extern int
 unmap_bus_range(struct pci_bus *bus);
 

include/linux/fsnotify.h

@@ -46,10 +46,8 @@ static inline void fsnotify_move(struct inode *old_dir, struct inode *new_dir,
 /*
  * fsnotify_unlink - file was unlinked
  */
-static inline void fsnotify_unlink(struct dentry *dentry, struct inode *dir)
+static inline void fsnotify_unlink(struct dentry *dentry, struct inode *inode, struct inode *dir)
 {
-	struct inode *inode = dentry->d_inode;
-
 	inode_dir_notify(dir, DN_DELETE);
 	inotify_inode_queue_event(dir, IN_DELETE, 0, dentry->d_name.name);
 	inotify_inode_queue_event(inode, IN_DELETE_SELF, 0, NULL);

include/linux/pci.h

@@ -225,6 +225,7 @@
 #define  PCI_PM_CAP_PME_D3cold  0x8000  /* PME# from D3 (cold) */
 #define PCI_PM_CTRL		4	/* PM control and status register */
 #define  PCI_PM_CTRL_STATE_MASK	0x0003	/* Current power state (D0 to D3) */
+#define  PCI_PM_CTRL_NO_SOFT_RESET	0x0004	/* No reset for D3hot->D0 */
 #define  PCI_PM_CTRL_PME_ENABLE	0x0100	/* PME pin enable */
 #define  PCI_PM_CTRL_DATA_SEL_MASK	0x1e00	/* Data select (??) */
 #define  PCI_PM_CTRL_DATA_SCALE_MASK	0x6000	/* Data scale (??) */
@@ -816,7 +817,9 @@ int pci_set_mwi(struct pci_dev *dev);
 void pci_clear_mwi(struct pci_dev *dev);
 int pci_set_dma_mask(struct pci_dev *dev, u64 mask);
 int pci_set_consistent_dma_mask(struct pci_dev *dev, u64 mask);
+void pci_update_resource(struct pci_dev *dev, struct resource *res, int resno);
 int pci_assign_resource(struct pci_dev *dev, int i);
+void pci_restore_bars(struct pci_dev *dev);
 
 /* ROM control related routines */
 void __iomem *pci_map_rom(struct pci_dev *pdev, size_t *size);

mm/mmap.c

@@ -143,7 +143,11 @@ int __vm_enough_memory(long pages, int cap_sys_admin)
 	   leave 3% of the size of this process for other processes */
 	allowed -= current->mm->total_vm / 32;
 
-	if (atomic_read(&vm_committed_space) < allowed)
+	/*
+	 * cast `allowed' as a signed long because vm_committed_space
+	 * sometimes has a negative value
+	 */
+	if (atomic_read(&vm_committed_space) < (long)allowed)
 		return 0;
 
 	vm_unacct_memory(pages);

mm/nommu.c

@@ -1167,7 +1167,11 @@ int __vm_enough_memory(long pages, int cap_sys_admin)
 	   leave 3% of the size of this process for other processes */
 	allowed -= current->mm->total_vm / 32;
 
-	if (atomic_read(&vm_committed_space) < allowed)
+	/*
+	 * cast `allowed' as a signed long because vm_committed_space
+	 * sometimes has a negative value
+	 */
+	if (atomic_read(&vm_committed_space) < (long)allowed)
 		return 0;
 
 	vm_unacct_memory(pages);

net/ipv4/tcp_output.c

@@ -403,11 +403,9 @@ static void tcp_queue_skb(struct sock *sk, struct sk_buff *skb)
 		sk->sk_send_head = skb;
 }
 
-static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb)
+static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb, unsigned int mss_now)
 {
-	struct tcp_sock *tp = tcp_sk(sk);
-
-	if (skb->len <= tp->mss_cache ||
+	if (skb->len <= mss_now ||
 	    !(sk->sk_route_caps & NETIF_F_TSO)) {
 		/* Avoid the costly divide in the normal
 		 * non-TSO case.
@@ -417,10 +415,10 @@ static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb)
 	} else {
 		unsigned int factor;
 
-		factor = skb->len + (tp->mss_cache - 1);
-		factor /= tp->mss_cache;
+		factor = skb->len + (mss_now - 1);
+		factor /= mss_now;
 		skb_shinfo(skb)->tso_segs = factor;
-		skb_shinfo(skb)->tso_size = tp->mss_cache;
+		skb_shinfo(skb)->tso_size = mss_now;
 	}
 }
 
@@ -429,7 +427,7 @@ static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb)
  * packet to the list.  This won't be called frequently, I hope. 
  * Remember, these are still headerless SKBs at this point.
  */
-static int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len)
+static int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len, unsigned int mss_now)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct sk_buff *buff;
@@ -492,8 +490,8 @@ static int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len)
 	}
 
 	/* Fix up tso_factor for both original and new SKB.  */
-	tcp_set_skb_tso_segs(sk, skb);
-	tcp_set_skb_tso_segs(sk, buff);
+	tcp_set_skb_tso_segs(sk, skb, mss_now);
+	tcp_set_skb_tso_segs(sk, buff, mss_now);
 
 	if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST) {
 		tp->lost_out += tcp_skb_pcount(skb);
@@ -569,7 +567,7 @@ int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len)
 	 * factor and mss.
 	 */
 	if (tcp_skb_pcount(skb) > 1)
-		tcp_set_skb_tso_segs(sk, skb);
+		tcp_set_skb_tso_segs(sk, skb, tcp_current_mss(sk, 1));
 
 	return 0;
 }
@@ -734,12 +732,14 @@ static inline unsigned int tcp_cwnd_test(struct tcp_sock *tp, struct sk_buff *sk
 /* This must be invoked the first time we consider transmitting
  * SKB onto the wire.
  */
-static inline int tcp_init_tso_segs(struct sock *sk, struct sk_buff *skb)
+static inline int tcp_init_tso_segs(struct sock *sk, struct sk_buff *skb, unsigned int mss_now)
 {
 	int tso_segs = tcp_skb_pcount(skb);
 
-	if (!tso_segs) {
-		tcp_set_skb_tso_segs(sk, skb);
+	if (!tso_segs ||
+	    (tso_segs > 1 &&
+	     skb_shinfo(skb)->tso_size != mss_now)) {
+		tcp_set_skb_tso_segs(sk, skb, mss_now);
 		tso_segs = tcp_skb_pcount(skb);
 	}
 	return tso_segs;
@@ -817,7 +817,7 @@ static unsigned int tcp_snd_test(struct sock *sk, struct sk_buff *skb,
 	struct tcp_sock *tp = tcp_sk(sk);
 	unsigned int cwnd_quota;
 
-	tcp_init_tso_segs(sk, skb);
+	tcp_init_tso_segs(sk, skb, cur_mss);
 
 	if (!tcp_nagle_test(tp, skb, cur_mss, nonagle))
 		return 0;
@@ -854,7 +854,7 @@ int tcp_may_send_now(struct sock *sk, struct tcp_sock *tp)
  * know that all the data is in scatter-gather pages, and that the
  * packet has never been sent out before (and thus is not cloned).
  */
-static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len)
+static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len, unsigned int mss_now)
 {
 	struct sk_buff *buff;
 	int nlen = skb->len - len;
@@ -887,8 +887,8 @@ static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len)
 	skb_split(skb, buff, len);
 
 	/* Fix up tso_factor for both original and new SKB.  */
-	tcp_set_skb_tso_segs(sk, skb);
-	tcp_set_skb_tso_segs(sk, buff);
+	tcp_set_skb_tso_segs(sk, skb, mss_now);
+	tcp_set_skb_tso_segs(sk, buff, mss_now);
 
 	/* Link BUFF into the send queue. */
 	skb_header_release(buff);
@@ -972,19 +972,18 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle)
 	if (unlikely(sk->sk_state == TCP_CLOSE))
 		return 0;
 
-	skb = sk->sk_send_head;
-	if (unlikely(!skb))
-		return 0;
-
-	tso_segs = tcp_init_tso_segs(sk, skb);
-	cwnd_quota = tcp_cwnd_test(tp, skb);
-	if (unlikely(!cwnd_quota))
-		goto out;
-
 	sent_pkts = 0;
-	while (likely(tcp_snd_wnd_test(tp, skb, mss_now))) {
+	while ((skb = sk->sk_send_head)) {
+		tso_segs = tcp_init_tso_segs(sk, skb, mss_now);
 		BUG_ON(!tso_segs);
 
+		cwnd_quota = tcp_cwnd_test(tp, skb);
+		if (!cwnd_quota)
+			break;
+
+		if (unlikely(!tcp_snd_wnd_test(tp, skb, mss_now)))
+			break;
+
 		if (tso_segs == 1) {
 			if (unlikely(!tcp_nagle_test(tp, skb, mss_now,
 						     (tcp_skb_is_last(sk, skb) ?
@@ -1006,11 +1005,11 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle)
 					limit = skb->len - trim;
 			}
 			if (skb->len > limit) {
-				if (tso_fragment(sk, skb, limit))
+				if (tso_fragment(sk, skb, limit, mss_now))
 					break;
 			}
 		} else if (unlikely(skb->len > mss_now)) {
-			if (unlikely(tcp_fragment(sk, skb,  mss_now)))
+			if (unlikely(tcp_fragment(sk, skb,  mss_now, mss_now)))
 				break;
 		}
 
@@ -1026,27 +1025,12 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle)
 
 		tcp_minshall_update(tp, mss_now, skb);
 		sent_pkts++;
-
-		/* Do not optimize this to use tso_segs. If we chopped up
-		 * the packet above, tso_segs will no longer be valid.
-		 */
-		cwnd_quota -= tcp_skb_pcount(skb);
-
-		BUG_ON(cwnd_quota < 0);
-		if (!cwnd_quota)
-			break;
-
-		skb = sk->sk_send_head;
-		if (!skb)
-			break;
-		tso_segs = tcp_init_tso_segs(sk, skb);
 	}
 
 	if (likely(sent_pkts)) {
 		tcp_cwnd_validate(sk, tp);
 		return 0;
 	}
-out:
 	return !tp->packets_out && sk->sk_send_head;
 }
 
@@ -1076,7 +1060,7 @@ void tcp_push_one(struct sock *sk, unsigned int mss_now)
 
 	BUG_ON(!skb || skb->len < mss_now);
 
-	tso_segs = tcp_init_tso_segs(sk, skb);
+	tso_segs = tcp_init_tso_segs(sk, skb, mss_now);
 	cwnd_quota = tcp_snd_test(sk, skb, mss_now, TCP_NAGLE_PUSH);
 
 	if (likely(cwnd_quota)) {
@@ -1093,11 +1077,11 @@ void tcp_push_one(struct sock *sk, unsigned int mss_now)
 					limit = skb->len - trim;
 			}
 			if (skb->len > limit) {
-				if (unlikely(tso_fragment(sk, skb, limit)))
+				if (unlikely(tso_fragment(sk, skb, limit, mss_now)))
 					return;
 			}
 		} else if (unlikely(skb->len > mss_now)) {
-			if (unlikely(tcp_fragment(sk, skb, mss_now)))
+			if (unlikely(tcp_fragment(sk, skb, mss_now, mss_now)))
 				return;
 		}
 
@@ -1388,7 +1372,7 @@ int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
 		int old_factor = tcp_skb_pcount(skb);
 		int new_factor;
 
-		if (tcp_fragment(sk, skb, cur_mss))
+		if (tcp_fragment(sk, skb, cur_mss, cur_mss))
 			return -ENOMEM; /* We'll try again later. */
 
 		/* New SKB created, account for it. */
@@ -1991,7 +1975,7 @@ int tcp_write_wakeup(struct sock *sk)
 			    skb->len > mss) {
 				seg_size = min(seg_size, mss);
 				TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
-				if (tcp_fragment(sk, skb, seg_size))
+				if (tcp_fragment(sk, skb, seg_size, mss))
 					return -1;
 				/* SWS override triggered forced fragmentation.
 				 * Disable TSO, the connection is too sick. */
@@ -2000,7 +1984,7 @@ int tcp_write_wakeup(struct sock *sk)
 					sk->sk_route_caps &= ~NETIF_F_TSO;
 				}
 			} else if (!tcp_skb_pcount(skb))
-				tcp_set_skb_tso_segs(sk, skb);
+				tcp_set_skb_tso_segs(sk, skb, mss);
 
 			TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
 			TCP_SKB_CB(skb)->when = tcp_time_stamp;