diff options
Diffstat (limited to 'Documentation')
| -rw-r--r-- | Documentation/00-INDEX | 2 | ||||
| -rw-r--r-- | Documentation/PCI/MSI-HOWTO.txt | 89 | ||||
| -rw-r--r-- | Documentation/SubmittingDrivers | 2 | ||||
| -rw-r--r-- | Documentation/SubmittingPatches | 2 | ||||
| -rw-r--r-- | Documentation/block/cfq-iosched.txt | 71 | ||||
| -rw-r--r-- | Documentation/email-clients.txt | 12 | ||||
| -rw-r--r-- | Documentation/filesystems/befs.txt | 2 | ||||
| -rw-r--r-- | Documentation/kernel-docs.txt | 11 | ||||
| -rw-r--r-- | Documentation/kernel-parameters.txt | 82 | ||||
| -rw-r--r-- | Documentation/networking/00-INDEX | 116 | ||||
| -rw-r--r-- | Documentation/networking/bonding.txt | 29 | ||||
| -rw-r--r-- | Documentation/networking/ip-sysctl.txt | 2 | ||||
| -rw-r--r-- | Documentation/networking/scaling.txt | 378 | ||||
| -rw-r--r-- | Documentation/ramoops.txt | 76 | ||||
| -rw-r--r-- | Documentation/virtual/00-INDEX | 3 | ||||
| -rw-r--r-- | Documentation/virtual/lguest/lguest.c | 3 | ||||
| -rw-r--r-- | Documentation/virtual/virtio-spec.txt | 2200 |
17 files changed, 2990 insertions, 90 deletions
diff --git a/Documentation/00-INDEX b/Documentation/00-INDEX index 1f89424c36a6..65bbd2622396 100644 --- a/Documentation/00-INDEX +++ b/Documentation/00-INDEX @@ -272,6 +272,8 @@ printk-formats.txt - how to get printk format specifiers right prio_tree.txt - info on radix-priority-search-tree use for indexing vmas. +ramoops.txt + - documentation of the ramoops oops/panic logging module. rbtree.txt - info on what red-black trees are and what they are for. robust-futex-ABI.txt diff --git a/Documentation/PCI/MSI-HOWTO.txt b/Documentation/PCI/MSI-HOWTO.txt index 3f5e0b09bed5..53e6fca146d7 100644 --- a/Documentation/PCI/MSI-HOWTO.txt +++ b/Documentation/PCI/MSI-HOWTO.txt @@ -45,7 +45,7 @@ arrived in memory (this becomes more likely with devices behind PCI-PCI bridges). In order to ensure that all the data has arrived in memory, the interrupt handler must read a register on the device which raised the interrupt. PCI transaction ordering rules require that all the data -arrives in memory before the value can be returned from the register. +arrive in memory before the value may be returned from the register. Using MSIs avoids this problem as the interrupt-generating write cannot pass the data writes, so by the time the interrupt is raised, the driver knows that all the data has arrived in memory. @@ -86,13 +86,13 @@ device. int pci_enable_msi(struct pci_dev *dev) -A successful call will allocate ONE interrupt to the device, regardless -of how many MSIs the device supports. The device will be switched from +A successful call allocates ONE interrupt to the device, regardless +of how many MSIs the device supports. The device is switched from pin-based interrupt mode to MSI mode. The dev->irq number is changed -to a new number which represents the message signaled interrupt. -This function should be called before the driver calls request_irq() -since enabling MSIs disables the pin-based IRQ and the driver will not -receive interrupts on the old interrupt. +to a new number which represents the message signaled interrupt; +consequently, this function should be called before the driver calls +request_irq(), because an MSI is delivered via a vector that is +different from the vector of a pin-based interrupt. 4.2.2 pci_enable_msi_block @@ -111,20 +111,20 @@ the device are in the range dev->irq to dev->irq + count - 1. If this function returns a negative number, it indicates an error and the driver should not attempt to request any more MSI interrupts for -this device. If this function returns a positive number, it will be -less than 'count' and indicate the number of interrupts that could have -been allocated. In neither case will the irq value have been -updated, nor will the device have been switched into MSI mode. +this device. If this function returns a positive number, it is +less than 'count' and indicates the number of interrupts that could have +been allocated. In neither case is the irq value updated or the device +switched into MSI mode. The device driver must decide what action to take if -pci_enable_msi_block() returns a value less than the number asked for. -Some devices can make use of fewer interrupts than the maximum they -request; in this case the driver should call pci_enable_msi_block() +pci_enable_msi_block() returns a value less than the number requested. +For instance, the driver could still make use of fewer interrupts; +in this case the driver should call pci_enable_msi_block() again. Note that it is not guaranteed to succeed, even when the 'count' has been reduced to the value returned from a previous call to pci_enable_msi_block(). This is because there are multiple constraints on the number of vectors that can be allocated; pci_enable_msi_block() -will return as soon as it finds any constraint that doesn't allow the +returns as soon as it finds any constraint that doesn't allow the call to succeed. 4.2.3 pci_disable_msi @@ -137,10 +137,10 @@ interrupt number and frees the previously allocated message signaled interrupt(s). The interrupt may subsequently be assigned to another device, so drivers should not cache the value of dev->irq. -A device driver must always call free_irq() on the interrupt(s) -for which it has called request_irq() before calling this function. -Failure to do so will result in a BUG_ON(), the device will be left with -MSI enabled and will leak its vector. +Before calling this function, a device driver must always call free_irq() +on any interrupt for which it previously called request_irq(). +Failure to do so results in a BUG_ON(), leaving the device with +MSI enabled and thus leaking its vector. 4.3 Using MSI-X @@ -155,10 +155,10 @@ struct msix_entry { }; This allows for the device to use these interrupts in a sparse fashion; -for example it could use interrupts 3 and 1027 and allocate only a +for example, it could use interrupts 3 and 1027 and yet allocate only a two-element array. The driver is expected to fill in the 'entry' value -in each element of the array to indicate which entries it wants the kernel -to assign interrupts for. It is invalid to fill in two entries with the +in each element of the array to indicate for which entries the kernel +should assign interrupts; it is invalid to fill in two entries with the same number. 4.3.1 pci_enable_msix @@ -168,10 +168,11 @@ int pci_enable_msix(struct pci_dev *dev, struct msix_entry *entries, int nvec) Calling this function asks the PCI subsystem to allocate 'nvec' MSIs. The 'entries' argument is a pointer to an array of msix_entry structs which should be at least 'nvec' entries in size. On success, the -function will return 0 and the device will have been switched into -MSI-X interrupt mode. The 'vector' elements in each entry will have -been filled in with the interrupt number. The driver should then call -request_irq() for each 'vector' that it decides to use. +device is switched into MSI-X mode and the function returns 0. +The 'vector' member in each entry is populated with the interrupt number; +the driver should then call request_irq() for each 'vector' that it +decides to use. The device driver is responsible for keeping track of the +interrupts assigned to the MSI-X vectors so it can free them again later. If this function returns a negative number, it indicates an error and the driver should not attempt to allocate any more MSI-X interrupts for @@ -181,16 +182,14 @@ below. This function, in contrast with pci_enable_msi(), does not adjust dev->irq. The device will not generate interrupts for this interrupt -number once MSI-X is enabled. The device driver is responsible for -keeping track of the interrupts assigned to the MSI-X vectors so it can -free them again later. +number once MSI-X is enabled. Device drivers should normally call this function once per device during the initialization phase. -It is ideal if drivers can cope with a variable number of MSI-X interrupts, +It is ideal if drivers can cope with a variable number of MSI-X interrupts; there are many reasons why the platform may not be able to provide the -exact number a driver asks for. +exact number that a driver asks for. A request loop to achieve that might look like: @@ -212,15 +211,15 @@ static int foo_driver_enable_msix(struct foo_adapter *adapter, int nvec) void pci_disable_msix(struct pci_dev *dev) -This API should be used to undo the effect of pci_enable_msix(). It frees +This function should be used to undo the effect of pci_enable_msix(). It frees the previously allocated message signaled interrupts. The interrupts may subsequently be assigned to another device, so drivers should not cache the value of the 'vector' elements over a call to pci_disable_msix(). -A device driver must always call free_irq() on the interrupt(s) -for which it has called request_irq() before calling this function. -Failure to do so will result in a BUG_ON(), the device will be left with -MSI enabled and will leak its vector. +Before calling this function, a device driver must always call free_irq() +on any interrupt for which it previously called request_irq(). +Failure to do so results in a BUG_ON(), leaving the device with +MSI-X enabled and thus leaking its vector. 4.3.3 The MSI-X Table @@ -232,10 +231,10 @@ mask or unmask an interrupt, it should call disable_irq() / enable_irq(). 4.4 Handling devices implementing both MSI and MSI-X capabilities If a device implements both MSI and MSI-X capabilities, it can -run in either MSI mode or MSI-X mode but not both simultaneously. +run in either MSI mode or MSI-X mode, but not both simultaneously. This is a requirement of the PCI spec, and it is enforced by the PCI layer. Calling pci_enable_msi() when MSI-X is already enabled or -pci_enable_msix() when MSI is already enabled will result in an error. +pci_enable_msix() when MSI is already enabled results in an error. If a device driver wishes to switch between MSI and MSI-X at runtime, it must first quiesce the device, then switch it back to pin-interrupt mode, before calling pci_enable_msi() or pci_enable_msix() and resuming @@ -251,7 +250,7 @@ the MSI-X facilities in preference to the MSI facilities. As mentioned above, MSI-X supports any number of interrupts between 1 and 2048. In constrast, MSI is restricted to a maximum of 32 interrupts (and must be a power of two). In addition, the MSI interrupt vectors must -be allocated consecutively, so the system may not be able to allocate +be allocated consecutively, so the system might not be able to allocate as many vectors for MSI as it could for MSI-X. On some platforms, MSI interrupts must all be targeted at the same set of CPUs whereas MSI-X interrupts can all be targeted at different CPUs. @@ -281,7 +280,7 @@ disabled to enabled and back again. Using 'lspci -v' (as root) may show some devices with "MSI", "Message Signalled Interrupts" or "MSI-X" capabilities. Each of these capabilities -has an 'Enable' flag which will be followed with either "+" (enabled) +has an 'Enable' flag which is followed with either "+" (enabled) or "-" (disabled). @@ -298,7 +297,7 @@ The PCI stack provides three ways to disable MSIs: Some host chipsets simply don't support MSIs properly. If we're lucky, the manufacturer knows this and has indicated it in the ACPI -FADT table. In this case, Linux will automatically disable MSIs. +FADT table. In this case, Linux automatically disables MSIs. Some boards don't include this information in the table and so we have to detect them ourselves. The complete list of these is found near the quirk_disable_all_msi() function in drivers/pci/quirks.c. @@ -317,7 +316,7 @@ Some bridges allow you to enable MSIs by changing some bits in their PCI configuration space (especially the Hypertransport chipsets such as the nVidia nForce and Serverworks HT2000). As with host chipsets, Linux mostly knows about them and automatically enables MSIs if it can. -If you have a bridge which Linux doesn't yet know about, you can enable +If you have a bridge unknown to Linux, you can enable MSIs in configuration space using whatever method you know works, then enable MSIs on that bridge by doing: @@ -327,7 +326,7 @@ where $bridge is the PCI address of the bridge you've enabled (eg 0000:00:0e.0). To disable MSIs, echo 0 instead of 1. Changing this value should be -done with caution as it can break interrupt handling for all devices +done with caution as it could break interrupt handling for all devices below this bridge. Again, please notify linux-pci@vger.kernel.org of any bridges that need @@ -336,7 +335,7 @@ special handling. 5.3. Disabling MSIs on a single device Some devices are known to have faulty MSI implementations. Usually this -is handled in the individual device driver but occasionally it's necessary +is handled in the individual device driver, but occasionally it's necessary to handle this with a quirk. Some drivers have an option to disable use of MSI. While this is a convenient workaround for the driver author, it is not good practise, and should not be emulated. @@ -350,7 +349,7 @@ for your machine. You should also check your .config to be sure you have enabled CONFIG_PCI_MSI. Then, 'lspci -t' gives the list of bridges above a device. Reading -/sys/bus/pci/devices/*/msi_bus will tell you whether MSI are enabled (1) +/sys/bus/pci/devices/*/msi_bus will tell you whether MSIs are enabled (1) or disabled (0). If 0 is found in any of the msi_bus files belonging to bridges between the PCI root and the device, MSIs are disabled. diff --git a/Documentation/SubmittingDrivers b/Documentation/SubmittingDrivers index 319baa8b60dd..36d16bbf72c6 100644 --- a/Documentation/SubmittingDrivers +++ b/Documentation/SubmittingDrivers @@ -130,7 +130,7 @@ Linux kernel master tree: ftp.??.kernel.org:/pub/linux/kernel/... ?? == your country code, such as "us", "uk", "fr", etc. - http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git + http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git Linux kernel mailing list: linux-kernel@vger.kernel.org diff --git a/Documentation/SubmittingPatches b/Documentation/SubmittingPatches index 569f3532e138..4468ce24427c 100644 --- a/Documentation/SubmittingPatches +++ b/Documentation/SubmittingPatches @@ -303,7 +303,7 @@ patches that are being emailed around. The sign-off is a simple line at the end of the explanation for the patch, which certifies that you wrote it or otherwise have the right to -pass it on as a open-source patch. The rules are pretty simple: if you +pass it on as an open-source patch. The rules are pretty simple: if you can certify the below: Developer's Certificate of Origin 1.1 diff --git a/Documentation/block/cfq-iosched.txt b/Documentation/block/cfq-iosched.txt index e578feed6d81..6d670f570451 100644 --- a/Documentation/block/cfq-iosched.txt +++ b/Documentation/block/cfq-iosched.txt @@ -43,3 +43,74 @@ If one sets slice_idle=0 and if storage supports NCQ, CFQ internally switches to IOPS mode and starts providing fairness in terms of number of requests dispatched. Note that this mode switching takes effect only for group scheduling. For non-cgroup users nothing should change. + +CFQ IO scheduler Idling Theory +=============================== +Idling on a queue is primarily about waiting for the next request to come +on same queue after completion of a request. In this process CFQ will not +dispatch requests from other cfq queues even if requests are pending there. + +The rationale behind idling is that it can cut down on number of seeks +on rotational media. For example, if a process is doing dependent +sequential reads (next read will come on only after completion of previous +one), then not dispatching request from other queue should help as we +did not move the disk head and kept on dispatching sequential IO from +one queue. + +CFQ has following service trees and various queues are put on these trees. + + sync-idle sync-noidle async + +All cfq queues doing synchronous sequential IO go on to sync-idle tree. +On this tree we idle on each queue individually. + +All synchronous non-sequential queues go on sync-noidle tree. Also any +request which are marked with REQ_NOIDLE go on this service tree. On this +tree we do not idle on individual queues instead idle on the whole group +of queues or the tree. So if there are 4 queues waiting for IO to dispatch +we will idle only once last queue has dispatched the IO and there is +no more IO on this service tree. + +All async writes go on async service tree. There is no idling on async +queues. + +CFQ has some optimizations for SSDs and if it detects a non-rotational +media which can support higher queue depth (multiple requests at in +flight at a time), then it cuts down on idling of individual queues and +all the queues move to sync-noidle tree and only tree idle remains. This +tree idling provides isolation with buffered write queues on async tree. + +FAQ +=== +Q1. Why to idle at all on queues marked with REQ_NOIDLE. + +A1. We only do tree idle (all queues on sync-noidle tree) on queues marked + with REQ_NOIDLE. This helps in providing isolation with all the sync-idle + queues. Otherwise in presence of many sequential readers, other + synchronous IO might not get fair share of disk. + + For example, if there are 10 sequential readers doing IO and they get + 100ms each. If a REQ_NOIDLE request comes in, it will be scheduled + roughly after 1 second. If after completion of REQ_NOIDLE request we + do not idle, and after a couple of milli seconds a another REQ_NOIDLE + request comes in, again it will be scheduled after 1second. Repeat it + and notice how a workload can lose its disk share and suffer due to + multiple sequential readers. + + fsync can generate dependent IO where bunch of data is written in the + context of fsync, and later some journaling data is written. Journaling + data comes in only after fsync has finished its IO (atleast for ext4 + that seemed to be the case). Now if one decides not to idle on fsync + thread due to REQ_NOIDLE, then next journaling write will not get + scheduled for another second. A process doing small fsync, will suffer + badly in presence of multiple sequential readers. + + Hence doing tree idling on threads using REQ_NOIDLE flag on requests + provides isolation from multiple sequential readers and at the same + time we do not idle on individual threads. + +Q2. When to specify REQ_NOIDLE +A2. I would think whenever one is doing synchronous write and not expecting + more writes to be dispatched from same context soon, should be able + to specify REQ_NOIDLE on writes and that probably should work well for + most of the cases. diff --git a/Documentation/email-clients.txt b/Documentation/email-clients.txt index a0b58e29f911..860c29a472ad 100644 --- a/Documentation/email-clients.txt +++ b/Documentation/email-clients.txt @@ -199,18 +199,16 @@ to coerce it into behaving. To beat some sense out of the internal editor, do this: -- Under account settings, composition and addressing, uncheck "Compose - messages in HTML format". - - Edit your Thunderbird config settings so that it won't use format=flowed. Go to "edit->preferences->advanced->config editor" to bring up the thunderbird's registry editor, and set "mailnews.send_plaintext_flowed" to "false". -- Enable "preformat" mode: Shft-click on the Write icon to bring up the HTML - composer, select "Preformat" from the drop-down box just under the subject - line, then close the message without saving. (This setting also applies to - the text composer, but the only control for it is in the HTML composer.) +- Disable HTML Format: Set "mail.identity.id1.compose_html" to "false". + +- Enable "preformat" mode: Set "editor.quotesPreformatted" to "true". + +- Enable UTF8: Set "prefs.converted-to-utf8" to "true". - Install the "toggle wordwrap" extension. Download the file from: https://addons.mozilla.org/thunderbird/addon/2351/ diff --git a/Documentation/filesystems/befs.txt b/Documentation/filesystems/befs.txt index 6e49c363938e..da45e6c842b8 100644 --- a/Documentation/filesystems/befs.txt +++ b/Documentation/filesystems/befs.txt @@ -27,7 +27,7 @@ His original code can still be found at: Does anyone know of a more current email address for Makoto? He doesn't respond to the address given above... -Current maintainer: Sergey S. Kostyliov <rathamahata@php4.ru> +This filesystem doesn't have a maintainer. WHAT IS THIS DRIVER? ================== diff --git a/Documentation/kernel-docs.txt b/Documentation/kernel-docs.txt index 9a8674629a07..0e0734b509d8 100644 --- a/Documentation/kernel-docs.txt +++ b/Documentation/kernel-docs.txt @@ -620,17 +620,6 @@ (including this document itself) have been moved there, and might be more up to date than the web version. - * Name: "Linux Source Driver" - URL: http://lsd.linux.cz - Keywords: Browsing source code. - Description: "Linux Source Driver (LSD) is an application, which - can make browsing source codes of Linux kernel easier than you can - imagine. You can select between multiple versions of kernel (e.g. - 0.01, 1.0.0, 2.0.33, 2.0.34pre13, 2.0.0, 2.1.101 etc.). With LSD - you can search Linux kernel (fulltext, macros, types, functions - and variables) and LSD can generate patches for you on the fly - (files, directories or kernel)". - * Name: "Linux Kernel Source Reference" Author: Thomas Graichen. URL: http://marc.info/?l=linux-kernel&m=96446640102205&w=4 diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt index e279b7242912..614d0382e2cb 100644 --- a/Documentation/kernel-parameters.txt +++ b/Documentation/kernel-parameters.txt @@ -40,6 +40,7 @@ parameter is applicable: ALSA ALSA sound support is enabled. APIC APIC support is enabled. APM Advanced Power Management support is enabled. + ARM ARM architecture is enabled. AVR32 AVR32 architecture is enabled. AX25 Appropriate AX.25 support is enabled. BLACKFIN Blackfin architecture is enabled. @@ -49,6 +50,7 @@ parameter is applicable: EFI EFI Partitioning (GPT) is enabled EIDE EIDE/ATAPI support is enabled. FB The frame buffer device is enabled. + FTRACE Function tracing enabled. GCOV GCOV profiling is enabled. HW Appropriate hardware is enabled. IA-64 IA-64 architecture is enabled. @@ -69,6 +71,7 @@ parameter is applicable: Documentation/m68k/kernel-options.txt. MCA MCA bus support is enabled. MDA MDA console support is enabled. + MIPS MIPS architecture is enabled. MOUSE Appropriate mouse support is enabled. MSI Message Signaled Interrupts (PCI). MTD MTD (Memory Technology Device) support is enabled. @@ -100,7 +103,6 @@ parameter is applicable: SPARC Sparc architecture is enabled. SWSUSP Software suspend (hibernation) is enabled. SUSPEND System suspend states are enabled. - FTRACE Function tracing enabled. TPM TPM drivers are enabled. TS Appropriate touchscreen support is enabled. UMS USB Mass Storage support is enabled. @@ -115,7 +117,7 @@ parameter is applicable: X86-64 X86-64 architecture is enabled. More X86-64 boot options can be found in Documentation/x86/x86_64/boot-options.txt . - X86 Either 32bit or 64bit x86 (same as X86-32+X86-64) + X86 Either 32-bit or 64-bit x86 (same as X86-32+X86-64) XEN Xen support is enabled In addition, the following text indicates that the option: @@ -376,7 +378,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. atkbd.softrepeat= [HW] Use software keyboard repeat - autotest [IA64] + autotest [IA-64] baycom_epp= [HW,AX25] Format: <io>,<mode> @@ -681,8 +683,8 @@ bytes respectively. Such letter suffixes can also be entirely omitted. uart[8250],mmio32,<addr>[,options] Start an early, polled-mode console on the 8250/16550 UART at the specified I/O port or MMIO address. - MMIO inter-register address stride is either 8bit (mmio) - or 32bit (mmio32). + MMIO inter-register address stride is either 8-bit + (mmio) or 32-bit (mmio32). The options are the same as for ttyS, above. earlyprintk= [X86,SH,BLACKFIN] @@ -725,7 +727,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. See Documentation/block/as-iosched.txt and Documentation/block/deadline-iosched.txt for details. - elfcorehdr= [IA64,PPC,SH,X86] + elfcorehdr= [IA-64,PPC,SH,X86] Specifies physical address of start of kernel core image elf header. Generally kexec loader will pass this option to capture kernel. @@ -791,7 +793,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. tracer at boot up. function-list is a comma separated list of functions. This list can be changed at run time by the set_ftrace_filter file in the debugfs - tracing directory. + tracing directory. ftrace_notrace=[function-list] [FTRACE] Do not trace the functions specified in @@ -829,7 +831,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. hashdist= [KNL,NUMA] Large hashes allocated during boot are distributed across NUMA nodes. Defaults on - for 64bit NUMA, off otherwise. + for 64-bit NUMA, off otherwise. Format: 0 | 1 (for off | on) hcl= [IA-64] SGI's Hardware Graph compatibility layer @@ -998,10 +1000,10 @@ bytes respectively. Such letter suffixes can also be entirely omitted. DMA. forcedac [x86_64] With this option iommu will not optimize to look - for io virtual address below 32 bit forcing dual + for io virtual address below 32-bit forcing dual address cycle on pci bus for cards supporting greater - than 32 bit addressing. The default is to look - for translation below 32 bit and if not available + than 32-bit addressing. The default is to look + for translation below 32-bit and if not available then look in the higher range. strict [Default Off] With this option on every unmap_single operation will @@ -1017,7 +1019,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. off disable Interrupt Remapping nosid disable Source ID checking - inttest= [IA64] + inttest= [IA-64] iomem= Disable strict checking of access to MMIO memory strict regions from userspace. @@ -1034,7 +1036,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. nomerge forcesac soft - pt [x86, IA64] + pt [x86, IA-64] io7= [HW] IO7 for Marvel based alpha systems See comment before marvel_specify_io7 in @@ -1165,7 +1167,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. kvm-amd.npt= [KVM,AMD] Disable nested paging (virtualized MMU) for all guests. - Default is 1 (enabled) if in 64bit or 32bit-PAE mode + Default is 1 (enabled) if in 64-bit or 32-bit PAE mode. kvm-intel.ept= [KVM,Intel] Disable extended page tables (virtualized MMU) support on capable Intel chips. @@ -1202,10 +1204,10 @@ bytes respectively. Such letter suffixes can also be entirely omitted. libata.dma=0 Disable all PATA and SATA DMA libata.dma=1 PATA and SATA Disk DMA only libata.dma=2 ATAPI (CDROM) DMA only - libata.dma=4 Compact Flash DMA only + libata.dma=4 Compact Flash DMA only Combinations also work, so libata.dma=3 enables DMA for disks and CDROMs, but not CFs. - + libata.ignore_hpa= [LIBATA] Ignore HPA limit libata.ignore_hpa=0 keep BIOS limits (default) libata.ignore_hpa=1 ignore limits, using full disk @@ -1331,7 +1333,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. ltpc= [NET] Format: <io>,<irq>,<dma> - machvec= [IA64] Force the use of a particular machine-vector + machvec= [IA-64] Force the use of a particular machine-vector (machvec) in a generic kernel. Example: machvec=hpzx1_swiotlb @@ -1348,9 +1350,12 @@ bytes respectively. Such letter suffixes can also be entirely omitted. it is equivalent to "nosmp", which also disables the IO APIC. - max_loop= [LOOP] Maximum number of loopback devices that can - be mounted - Format: <1-256> + max_loop= [LOOP] The number of loop block devices that get + (loop.max_loop) unconditionally pre-created at init time. The default + number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead + of statically allocating a predefined number, loop + devices can be requested on-demand with the + /dev/loop-control interface. mcatest= [IA-64] @@ -1734,7 +1739,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. nointroute [IA-64] - nojitter [IA64] Disables jitter checking for ITC timers. + nojitter [IA-64] Disables jitter checking for ITC timers. no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver @@ -1800,7 +1805,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. nox2apic [X86-64,APIC] Do not enable x2APIC mode. - nptcg= [IA64] Override max number of concurrent global TLB + nptcg= [IA-64] Override max number of concurrent global TLB purges which is reported from either PAL_VM_SUMMARY or SAL PALO. @@ -2077,7 +2082,7 @@ bytes respectively. Such letter suffixes can also be entirely omitted. Format: { parport<nr> | timid | 0 } See also Documentation/parport.txt. - pmtmr= [X86] Manual setup of pmtmr I/O Port. + pmtmr= [X86] Manual setup of pmtmr I/O Port. Override pmtimer IOPort with a hex value. e.g. pmtmr=0x508 @@ -2635,6 +2640,16 @@ bytes respectively. Such letter suffixes can also be entirely omitted. medium is write-protected). Example: quirks=0419:aaf5:rl,0421:0433:rc + user_debug= [KNL,ARM] + Format: <int> + See arch/arm/Kconfig.debug help text. + 1 - undefined instruction events + 2 - system calls + 4 - invalid data aborts + 8 - SIGSEGV faults + 16 - SIGBUS faults + Example: user_debug=31 + userpte= [X86] Flags controlling user PTE allocations. @@ -2680,6 +2695,27 @@ bytes respectively. Such letter suffixes can also be entirely omitted. vmpoff= [KNL,S390] Perform z/VM CP command after power off. Format: <command> + vsyscall= [X86-64] + Controls the behavior of vsyscalls (i.e. calls to + fixed addresses of 0xffffffffff600x00 from legacy + code). Most statically-linked binaries and older + versions of glibc use these calls. Because these + functions are at fixed addresses, they make nice + targets for exploits that can control RIP. + + emulate [default] Vsyscalls turn into traps and are + emulated reasonably safely. + + native Vsyscalls are native syscall instructions. + This is a little bit faster than trapping + and makes a few dynamic recompilers work + better than they would in emulation mode. + It also makes exploits much easier to write. + + none Vsyscalls don't work at all. This makes + them quite hard to use for exploits but + might break your system. + vt.cur_default= [VT] Default cursor shape. Format: 0xCCBBAA, where AA, BB, and CC are the same as the parameters of the <Esc>[?A;B;Cc escape sequence; diff --git a/Documentation/networking/00-INDEX b/Documentation/networking/00-INDEX index 4edd78dfb362..bbce1215434a 100644 --- a/Documentation/networking/00-INDEX +++ b/Documentation/networking/00-INDEX @@ -1,13 +1,21 @@ 00-INDEX - this file +3c359.txt + - information on the 3Com TokenLink Velocity XL (3c5359) driver. 3c505.txt - information on the 3Com EtherLink Plus (3c505) driver. +3c509.txt + - information on the 3Com Etherlink III Series Ethernet cards. 6pack.txt - info on the 6pack protocol, an alternative to KISS for AX.25 DLINK.txt - info on the D-Link DE-600/DE-620 parallel port pocket adapters PLIP.txt - PLIP: The Parallel Line Internet Protocol device driver +README.ipw2100 + - README for the Intel PRO/Wireless 2100 driver. +README.ipw2200 + - README for the Intel PRO/Wireless 2915ABG and 2200BG driver. README.sb1000 - info on General Instrument/NextLevel SURFboard1000 cable modem. alias.txt @@ -20,8 +28,12 @@ atm.txt - info on where to get ATM programs and support for Linux. ax25.txt - info on using AX.25 and NET/ROM code for Linux +batman-adv.txt + - B.A.T.M.A.N routing protocol on top of layer 2 Ethernet Frames. baycom.txt - info on the driver for Baycom style amateur radio modems +bonding.txt + - Linux Ethernet Bonding Driver HOWTO: link aggregation in Linux. bridge.txt - where to get user space programs for ethernet bridging with Linux. can.txt @@ -34,32 +46,60 @@ cxacru.txt - Conexant AccessRunner USB ADSL Modem cxacru-cf.py - Conexant AccessRunner USB ADSL Modem configuration file parser +cxgb.txt + - Release Notes for the Chelsio N210 Linux device driver. +dccp.txt + - the Datagram Congestion Control Protocol (DCCP) (RFC 4340..42). de4x5.txt - the Digital EtherWORKS DE4?? and DE5?? PCI Ethernet driver decnet.txt - info on using the DECnet networking layer in Linux. depca.txt - the Digital DEPCA/EtherWORKS DE1?? and DE2?? LANCE Ethernet driver +dl2k.txt + - README for D-Link DL2000-based Gigabit Ethernet Adapters (dl2k.ko). +dm9000.txt + - README for the Simtec DM9000 Network driver. dmfe.txt - info on the Davicom DM9102(A)/DM9132/DM9801 fast ethernet driver. +dns_resolver.txt + - The DNS resolver module allows kernel servies to make DNS queries. +driver.txt + - Softnet driver issues. e100.txt - info on Intel's EtherExpress PRO/100 line of 10/100 boards e1000.txt - info on Intel's E1000 line of gigabit ethernet boards +e1000e.txt + - README for the Intel Gigabit Ethernet Driver (e1000e). eql.txt - serial IP load balancing ewrk3.txt - the Digital EtherWORKS 3 DE203/4/5 Ethernet driver +fib_trie.txt + - Level Compressed Trie (LC-trie) notes: a structure for routing. filter.txt - Linux Socket Filtering fore200e.txt - FORE Systems PCA-200E/SBA-200E ATM NIC driver info. framerelay.txt - info on using Frame Relay/Data Link Connection Identifier (DLCI). +gen_stats.txt + - Generic networking statistics for netlink users. +generic_hdlc.txt + - The generic High Level Data Link Control (HDLC) layer. generic_netlink.txt - info on Generic Netlink +gianfar.txt + - Gianfar Ethernet Driver. ieee802154.txt - Linux IEEE 802.15.4 implementation, API and drivers +ifenslave.c + - Configure network interfaces for parallel routing (bonding). +igb.txt + - README for the Intel Gigabit Ethernet Driver (igb). +igbvf.txt + - README for the Intel Gigabit Ethernet Driver (igbvf). ip-sysctl.txt - /proc/sys/net/ipv4/* variables ip_dynaddr.txt @@ -68,41 +108,117 @@ ipddp.txt - AppleTalk-IP Decapsulation and AppleTalk-IP Encapsulation iphase.txt - Interphase PCI ATM (i)Chip IA Linux driver info. +ipv6.txt + - Options to the ipv6 kernel module. +ipvs-sysctl.txt + - Per-inode explanation of the /proc/sys/net/ipv4/vs interface. irda.txt - where to get IrDA (infrared) utilities and info for Linux. +ixgb.txt + - README for the Intel 10 Gigabit Ethernet Driver (ixgb). +ixgbe.txt + - README for the Intel 10 Gigabit Ethernet Driver (ixgbe). +ixgbevf.txt + - README for the Intel Virtual Function (VF) Driver (ixgbevf). +l2tp.txt + - User guide to the L2TP tunnel protocol. lapb-module.txt - programming information of the LAPB module. ltpc.txt - the Apple or Farallon LocalTalk PC card driver +mac80211-injection.txt + - HOWTO use packet injection with mac80211 multicast.txt - Behaviour of cards under Multicast +multiqueue.txt + - HOWTO for multiqueue network device support. +netconsole.txt + - The network console module netconsole.ko: configuration and notes. +netdev-features.txt + - Network interface features API description. netdevices.txt - info on network device driver functions exported to the kernel. +netif-msg.txt + - Design of the network interface message level setting (NETIF_MSG_*). +nfc.txt + - The Linux Near Field Communication (NFS) subsystem. olympic.txt - IBM PCI Pit/Pit-Phy/Olympic Token Ring driver info. +operstates.txt + - Overview of network interface operational states. +packet_mmap.txt + - User guide to memory mapped packet socket rings (PACKET_[RT]X_RING). +phonet.txt + - The Phonet packet protocol used in Nokia cellular modems. +phy.txt + - The PHY abstraction layer. +pktgen.txt + - User guide to the kernel packet generator (pktgen.ko). policy-routing.txt - IP policy-based routing +ppp_generic.txt + - Information about the generic PPP driver. +proc_net_tcp.txt + - Per inode overview of the /proc/net/tcp and /proc/net/tcp6 interfaces. +radiotap-headers.txt + - Background on radiotap headers. ray_cs.txt - Raylink Wireless LAN card driver info. +rds.txt + - Background on the reliable, ordered datagram delivery method RDS. +regulatory.txt + - Overview of the Linux wireless regulatory infrastructure. +rxrpc.txt + - Guide to the RxRPC protocol. +s2io.txt + - Release notes for Neterion Xframe I/II 10GbE driver. +scaling.txt + - Explanation of network scaling techniques: RSS, RPS, RFS, aRFS, XPS. +sctp.txt + - Notes on the Linux kernel implementation of the SCTP protocol. +secid.txt + - Explanation of the secid member in flow structures. skfp.txt - SysKonnect FDDI (SK-5xxx, Compaq Netelligent) driver info. smc9.txt - the driver for SMC's 9000 series of Ethernet cards smctr.txt - SMC TokenCard TokenRing Linux driver info. +spider-net.txt + - README for the Spidernet Driver (as found in PS3 / Cell BE). +stmmac.txt + - README for the STMicro Synopsys Ethernet driver. +tc-actions-env-rules.txt + - rules for traffic control (tc) actions. +timestamping.txt + - overview of network packet timestamping variants. tcp.txt - short blurb on how TCP output takes place. +tcp-thin.txt + - kernel tuning options for low rate 'thin' TCP streams. tlan.txt - ThunderLAN (Compaq Netelligent 10/100, Olicom OC-2xxx) driver info. tms380tr.txt - SysKonnect Token Ring ISA/PCI adapter driver info. +tproxy.txt + - Transparent proxy support user guide. tuntap.txt - TUN/TAP device driver, allowing user space Rx/Tx of packets. +udplite.txt + - UDP-Lite protocol (RFC 3828) introduction. vortex.txt - info on using 3Com Vortex (3c590, 3c592, 3c595, 3c597) Ethernet cards. +vxge.txt + - README for the Neterion X3100 PCIe Server Adapter. x25.txt - general info on X.25 development. x25-iface.txt - description of the X.25 Packet Layer to LAPB device interface. +xfrm_proc.txt + - description of the statistics package for XFRM. +xfrm_sync.txt + - sync patches for XFRM enable migration of an SA between hosts. +xfrm_sysctl.txt + - description of the XFRM configuration options. z8530drv.txt - info about Linux driver for Z8530 based HDLC cards for AX.25 diff --git a/Documentation/networking/bonding.txt b/Documentation/networking/bonding.txt index 5dd960d75174..91df678fb7f8 100644 --- a/Documentation/networking/bonding.txt +++ b/Documentation/networking/bonding.txt @@ -238,6 +238,18 @@ ad_select This option was added in bonding version 3.4.0. +all_slaves_active + + Specifies that duplicate frames (received on inactive ports) should be + dropped (0) or delivered (1). + + Normally, bonding will drop duplicate frames (received on inactive + ports), which is desirable for most users. But there are some times + it is nice to allow duplicate frames to be delivered. + + The default value is 0 (drop duplicate frames received on inactive + ports). + arp_interval Specifies the ARP link monitoring frequency in milliseconds. @@ -433,6 +445,23 @@ miimon determined. See the High Availability section for additional information. The default value is 0. +min_links + + Specifies the minimum number of links that must be active before + asserting carrier. It is similar to the Cisco EtherChannel min-links + feature. This allows setting the minimum number of member ports that + must be up (link-up state) before marking the bond device as up + (carrier on). This is useful for situations where higher level services + such as clustering want to ensure a minimum number of low bandwidth + links are active before switchover. This option only affect 802.3ad + mode. + + The default value is 0. This will cause carrier to be asserted (for + 802.3ad mode) whenever there is an active aggregator, regardless of the + number of available links in that aggregator. Note that, because an + aggregator cannot be active without at least one available link, + setting this option to 0 or to 1 has the exact same effect. + mode Specifies one of the bonding policies. The default is diff --git a/Documentation/networking/ip-sysctl.txt b/Documentation/networking/ip-sysctl.txt index db2a4067013c..81546990f41c 100644 --- a/Documentation/networking/ip-sysctl.txt +++ b/Documentation/networking/ip-sysctl.txt @@ -992,7 +992,7 @@ bindv6only - BOOLEAN TRUE: disable IPv4-mapped address feature FALSE: enable IPv4-mapped address feature - Default: FALSE (as specified in RFC2553bis) + Default: FALSE (as specified in RFC3493) IPv6 Fragmentation: diff --git a/Documentation/networking/scaling.txt b/Documentation/networking/scaling.txt new file mode 100644 index 000000000000..58fd7414e6c0 --- /dev/null +++ b/Documentation/networking/scaling.txt @@ -0,0 +1,378 @@ +Scaling in the Linux Networking Stack + + +Introduction +============ + +This document describes a set of complementary techniques in the Linux +networking stack to increase parallelism and improve performance for +multi-processor systems. + +The following technologies are described: + + RSS: Receive Side Scaling + RPS: Receive Packet Steering + RFS: Receive Flow Steering + Accelerated Receive Flow Steering + XPS: Transmit Packet Steering + + +RSS: Receive Side Scaling +========================= + +Contemporary NICs support multiple receive and transmit descriptor queues +(multi-queue). On reception, a NIC can send different packets to different +queues to distribute processing among CPUs. The NIC distributes packets by +applying a filter to each packet that assigns it to one of a small number +of logical flows. Packets for each flow are steered to a separate receive +queue, which in turn can be processed by separate CPUs. This mechanism is +generally known as “Receive-side Scaling” (RSS). The goal of RSS and +the other scaling techniques to increase performance uniformly. +Multi-queue distribution can also be used for traffic prioritization, but +that is not the focus of these techniques. + +The filter used in RSS is typically a hash function over the network +and/or transport layer headers-- for example, a 4-tuple hash over +IP addresses and TCP ports of a packet. The most common hardware +implementation of RSS uses a 128-entry indirection table where each entry +stores a queue number. The receive queue for a packet is determined +by masking out the low order seven bits of the computed hash for the +packet (usually a Toeplitz hash), taking this number as a key into the +indirection table and reading the corresponding value. + +Some advanced NICs allow steering packets to queues based on +programmable filters. For example, webserver bound TCP port 80 packets +can be directed to their own receive queue. Such “n-tuple” filters can +be configured from ethtool (--config-ntuple). + +==== RSS Configuration + +The driver for a multi-queue capable NIC typically provides a kernel +module parameter for specifying the number of hardware queues to +configure. In the bnx2x driver, for instance, this parameter is called +num_queues. A typical RSS configuration would be to have one receive queue +for each CPU if the device supports enough queues, or otherwise at least +one for each memory domain, where a memory domain is a set of CPUs that +share a particular memory level (L1, L2, NUMA node, etc.). + +The indirection table of an RSS device, which resolves a queue by masked +hash, is usually programmed by the driver at initialization. The +default mapping is to distribute the queues evenly in the table, but the +indirection table can be retrieved and modified at runtime using ethtool +commands (--show-rxfh-indir and --set-rxfh-indir). Modifying the +indirection table could be done to give different queues different +relative weights. + +== RSS IRQ Configuration + +Each receive queue has a separate IRQ associated with it. The NIC triggers +this to notify a CPU when new packets arrive on the given queue. The +signaling path for PCIe devices uses message signaled interrupts (MSI-X), +that can route each interrupt to a particular CPU. The active mapping +of queues to IRQs can be determined from /proc/interrupts. By default, +an IRQ may be handled on any CPU. Because a non-negligible part of packet +processing takes place in receive interrupt handling, it is advantageous +to spread receive interrupts between CPUs. To manually adjust the IRQ +affinity of each interrupt see Documentation/IRQ-affinity. Some systems +will be running irqbalance, a daemon that dynamically optimizes IRQ +assignments and as a result may override any manual settings. + +== Suggested Configuration + +RSS should be enabled when latency is a concern or whenever receive +interrupt processing forms a bottleneck. Spreading load between CPUs +decreases queue length. For low latency networking, the optimal setting +is to allocate as many queues as there are CPUs in the system (or the +NIC maximum, if lower). The most efficient high-rate configuration +is likely the one with the smallest number of receive queues where no +receive queue overflows due to a saturated CPU, because in default +mode with interrupt coalescing enabled, the aggregate number of +interrupts (and thus work) grows with each additional queue. + +Per-cpu load can be observed using the mpstat utility, but note that on +processors with hyperthreading (HT), each hyperthread is represented as +a separate CPU. For interrupt handling, HT has shown no benefit in +initial tests, so limit the number of queues to the number of CPU cores +in the system. + + +RPS: Receive Packet Steering +============================ + +Receive Packet Steering (RPS) is logically a software implementation of +RSS. Being in software, it is necessarily called later in the datapath. +Whereas RSS selects the queue and hence CPU that will run the hardware +interrupt handler, RPS selects the CPU to perform protocol processing +above the interrupt handler. This is accomplished by placing the packet +on the desired CPU’s backlog queue and waking up the CPU for processing. +RPS has some advantages over RSS: 1) it can be used with any NIC, +2) software filters can easily be added to hash over new protocols, +3) it does not increase hardware device interrupt rate (although it does +introduce inter-processor interrupts (IPIs)). + +RPS is called during bottom half of the receive interrupt handler, when +a driver sends a packet up the network stack with netif_rx() or +netif_receive_skb(). These call the get_rps_cpu() function, which +selects the queue that should process a packet. + +The first step in determining the target CPU for RPS is to calculate a +flow hash over the packet’s addresses or ports (2-tuple or 4-tuple hash +depending on the protocol). This serves as a consistent hash of the +associated flow of the packet. The hash is either provided by hardware +or will be computed in the stack. Capable hardware can pass the hash in +the receive descriptor for the packet; this would usually be the same +hash used for RSS (e.g. computed Toeplitz hash). The hash is saved in +skb->rx_hash and can be used elsewhere in the stack as a hash of the +packet’s flow. + +Each receive hardware queue has an associated list of CPUs to which +RPS may enqueue packets for processing. For each received packet, +an index into the list is computed from the flow hash modulo the size +of the list. The indexed CPU is the target for processing the packet, +and the packet is queued to the tail of that CPU’s backlog queue. At +the end of the bottom half routine, IPIs are sent to any CPUs for which +packets have been queued to their backlog queue. The IPI wakes backlog +processing on the remote CPU, and any queued packets are then processed +up the networking stack. + +==== RPS Configuration + +RPS requires a kernel compiled with the CONFIG_RPS kconfig symbol (on +by default for SMP). Even when compiled in, RPS remains disabled until +explicitly configured. The list of CPUs to which RPS may forward traffic +can be configured for each receive queue using a sysfs file entry: + + /sys/class/net/<dev>/queues/rx-<n>/rps_cpus + +This file implements a bitmap of CPUs. RPS is disabled when it is zero +(the default), in which case packets are processed on the interrupting +CPU. Documentation/IRQ-affinity.txt explains how CPUs are assigned to +the bitmap. + +== Suggested Configuration + +For a single queue device, a typical RPS configuration would be to set +the rps_cpus to the CPUs in the same memory domain of the interrupting +CPU. If NUMA locality is not an issue, this could also be all CPUs in +the system. At high interrupt rate, it might be wise to exclude the +interrupting CPU from the map since that already performs much work. + +For a multi-queue system, if RSS is configured so that a hardware +receive queue is mapped to each CPU, then RPS is probably redundant +and unnecessary. If there are fewer hardware queues than CPUs, then +RPS might be beneficial if the rps_cpus for each queue are the ones that +share the same memory domain as the interrupting CPU for that queue. + + +RFS: Receive Flow Steering +========================== + +While RPS steers packets solely based on hash, and thus generally +provides good load distribution, it does not take into account +application locality. This is accomplished by Receive Flow Steering +(RFS). The goal of RFS is to increase datacache hitrate by steering +kernel processing of packets to the CPU where the application thread +consuming the packet is running. RFS relies on the same RPS mechanisms +to enqueue packets onto the backlog of another CPU and to wake up that +CPU. + +In RFS, packets are not forwarded directly by the value of their hash, +but the hash is used as index into a flow lookup table. This table maps +flows to the CPUs where those flows are being processed. The flow hash +(see RPS section above) is used to calculate the index into this table. +The CPU recorded in each entry is the one which last processed the flow. +If an entry does not hold a valid CPU, then packets mapped to that entry +are steered using plain RPS. Multiple table entries may point to the +same CPU. Indeed, with many flows and few CPUs, it is very likely that +a single application thread handles flows with many different flow hashes. + +rps_sock_table is a global flow table that contains the *desired* CPU for +flows: the CPU that is currently processing the flow in userspace. Each +table value is a CPU index that is updated during calls to recvmsg and +sendmsg (specifically, inet_recvmsg(), inet_sendmsg(), inet_sendpage() +and tcp_splice_read()). + +When the scheduler moves a thread to a new CPU while it has outstanding +receive packets on the old CPU, packets may arrive out of order. To +avoid this, RFS uses a second flow table to track outstanding packets +for each flow: rps_dev_flow_table is a table specific to each hardware +receive queue of each device. Each table value stores a CPU index and a +counter. The CPU index represents the *current* CPU onto which packets +for this flow are enqueued for further kernel processing. Ideally, kernel +and userspace processing occur on the same CPU, and hence the CPU index +in both tables is identical. This is likely false if the scheduler has +recently migrated a userspace thread while the kernel still has packets +enqueued for kernel processing on the old CPU. + +The counter in rps_dev_flow_table values records the length of the current +CPU's backlog when a packet in this flow was last enqueued. Each backlog +queue has a head counter that is incremented on dequeue. A tail counter +is computed as head counter + queue length. In other words, the counter +in rps_dev_flow_table[i] records the last element in flow i that has +been enqueued onto the currently designated CPU for flow i (of course, +entry i is actually selected by hash and multiple flows may hash to the +same entry i). + +And now the trick for avoiding out of order packets: when selecting the +CPU for packet processing (from get_rps_cpu()) the rps_sock_flow table +and the rps_dev_flow table of the queue that the packet was received on +are compared. If the desired CPU for the flow (found in the +rps_sock_flow table) matches the current CPU (found in the rps_dev_flow +table), the packet is enqueued onto that CPU’s backlog. If they differ, +the current CPU is updated to match the desired CPU if one of the +following is true: + +- The current CPU's queue head counter >= the recorded tail counter + value in rps_dev_flow[i] +- The current CPU is unset (equal to NR_CPUS) +- The current CPU is offline + +After this check, the packet is sent to the (possibly updated) current +CPU. These rules aim to ensure that a flow only moves to a new CPU when +there are no packets outstanding on the old CPU, as the outstanding +packets could arrive later than those about to be processed on the new +CPU. + +==== RFS Configuration + +RFS is only available if the kconfig symbol CONFIG_RFS is enabled (on +by default for SMP). The functionality remains disabled until explicitly +configured. The number of entries in the global flow table is set through: + + /proc/sys/net/core/rps_sock_flow_entries + +The number of entries in the per-queue flow table are set through: + + /sys/class/net/<dev>/queues/tx-<n>/rps_flow_cnt + +== Suggested Configuration + +Both of these need to be set before RFS is enabled for a receive queue. +Values for both are rounded up to the nearest power of two. The +suggested flow count depends on the expected number of active connections +at any given time, which may be significantly less than the number of open +connections. We have found that a value of 32768 for rps_sock_flow_entries +works fairly well on a moderately loaded server. + +For a single queue device, the rps_flow_cnt value for the single queue +would normally be configured to the same value as rps_sock_flow_entries. +For a multi-queue device, the rps_flow_cnt for each queue might be +configured as rps_sock_flow_entries / N, where N is the number of +queues. So for instance, if rps_flow_entries is set to 32768 and there +are 16 configured receive queues, rps_flow_cnt for each queue might be +configured as 2048. + + +Accelerated RFS +=============== + +Accelerated RFS is to RFS what RSS is to RPS: a hardware-accelerated load +balancing mechanism that uses soft state to steer flows based on where +the application thread consuming the packets of each flow is running. +Accelerated RFS should perform better than RFS since packets are sent +directly to a CPU local to the thread consuming the data. The target CPU +will either be the same CPU where the application runs, or at least a CPU +which is local to the application thread’s CPU in the cache hierarchy. + +To enable accelerated RFS, the networking stack calls the +ndo_rx_flow_steer driver function to communicate the desired hardware +queue for packets matching a particular flow. The network stack +automatically calls this function every time a flow entry in +rps_dev_flow_table is updated. The driver in turn uses a device specific +method to program the NIC to steer the packets. + +The hardware queue for a flow is derived from the CPU recorded in +rps_dev_flow_table. The stack consults a CPU to hardware queue map which +is maintained by the NIC driver. This is an auto-generated reverse map of +the IRQ affinity table shown by /proc/interrupts. Drivers can use +functions in the cpu_rmap (“CPU affinity reverse map”) kernel library +to populate the map. For each CPU, the corresponding queue in the map is +set to be one whose processing CPU is closest in cache locality. + +==== Accelerated RFS Configuration + +Accelerated RFS is only available if the kernel is compiled with +CONFIG_RFS_ACCEL and support is provided by the NIC device and driver. +It also requires that ntuple filtering is enabled via ethtool. The map +of CPU to queues is automatically deduced from the IRQ affinities +configured for each receive queue by the driver, so no additional +configuration should be necessary. + +== Suggested Configuration + +This technique should be enabled whenever one wants to use RFS and the +NIC supports hardware acceleration. + +XPS: Transmit Packet Steering +============================= + +Transmit Packet Steering is a mechanism for intelligently selecting +which transmit queue to use when transmitting a packet on a multi-queue +device. To accomplish this, a mapping from CPU to hardware queue(s) is +recorded. The goal of this mapping is usually to assign queues +exclusively to a subset of CPUs, where the transmit completions for +these queues are processed on a CPU within this set. This choice +provides two benefits. First, contention on the device queue lock is +significantly reduced since fewer CPUs contend for the same queue +(contention can be eliminated completely if each CPU has its own +transmit queue). Secondly, cache miss rate on transmit completion is +reduced, in particular for data cache lines that hold the sk_buff +structures. + +XPS is configured per transmit queue by setting a bitmap of CPUs that +may use that queue to transmit. The reverse mapping, from CPUs to +transmit queues, is computed and maintained for each network device. +When transmitting the first packet in a flow, the function +get_xps_queue() is called to select a queue. This function uses the ID +of the running CPU as a key into the CPU-to-queue lookup table. If the +ID matches a single queue, that is used for transmission. If multiple +queues match, one is selected by using the flow hash to compute an index +into the set. + +The queue chosen for transmitting a particular flow is saved in the +corresponding socket structure for the flow (e.g. a TCP connection). +This transmit queue is used for subsequent packets sent on the flow to +prevent out of order (ooo) packets. The choice also amortizes the cost +of calling get_xps_queues() over all packets in the flow. To avoid +ooo packets, the queue for a flow can subsequently only be changed if +skb->ooo_okay is set for a packet in the flow. This flag indicates that +there are no outstanding packets in the flow, so the transmit queue can +change without the risk of generating out of order packets. The +transport layer is responsible for setting ooo_okay appropriately. TCP, +for instance, sets the flag when all data for a connection has been +acknowledged. + +==== XPS Configuration + +XPS is only available if the kconfig symbol CONFIG_XPS is enabled (on by +default for SMP). The functionality remains disabled until explicitly +configured. To enable XPS, the bitmap of CPUs that may use a transmit +queue is configured using the sysfs file entry: + +/sys/class/net/<dev>/queues/tx-<n>/xps_cpus + +== Suggested Configuration + +For a network device with a single transmission queue, XPS configuration +has no effect, since there is no choice in this case. In a multi-queue +system, XPS is preferably configured so that each CPU maps onto one queue. +If there are as many queues as there are CPUs in the system, then each +queue can also map onto one CPU, resulting in exclusive pairings that +experience no contention. If there are fewer queues than CPUs, then the +best CPUs to share a given queue are probably those that share the cache +with the CPU that processes transmit completions for that queue +(transmit interrupts). + + +Further Information +=================== +RPS and RFS were introduced in kernel 2.6.35. XPS was incorporated into +2.6.38. Original patches were submitted by Tom Herbert +(therbert@google.com) + +Accelerated RFS was introduced in 2.6.35. Original patches were +submitted by Ben Hutchings (bhutchings@solarflare.com) + +Authors: +Tom Herbert (therbert@google.com) +Willem de Bruijn (willemb@google.com) diff --git a/Documentation/ramoops.txt b/Documentation/ramoops.txt new file mode 100644 index 000000000000..8fb1ba7fe7bf --- /dev/null +++ b/Documentation/ramoops.txt @@ -0,0 +1,76 @@ +Ramoops oops/panic logger +========================= + +Sergiu Iordache <sergiu@chromium.org> + +Updated: 8 August 2011 + +0. Introduction + +Ramoops is an oops/panic logger that writes its logs to RAM before the system +crashes. It works by logging oopses and panics in a circular buffer. Ramoops +needs a system with persistent RAM so that the content of that area can +survive after a restart. + +1. Ramoops concepts + +Ramoops uses a predefined memory area to store the dump. The start and size of +the memory area are set using two variables: + * "mem_address" for the start + * "mem_size" for the size. The memory size will be rounded down to a + power of two. + +The memory area is divided into "record_size" chunks (also rounded down to +power of two) and each oops/panic writes a "record_size" chunk of +information. + +Dumping both oopses and panics can be done by setting 1 in the "dump_oops" +variable while setting 0 in that variable dumps only the panics. + +The module uses a counter to record multiple dumps but the counter gets reset +on restart (i.e. new dumps after the restart will overwrite old ones). + +2. Setting the parameters + +Setting the ramoops parameters can be done in 2 different manners: + 1. Use the module parameters (which have the names of the variables described + as before). + 2. Use a platform device and set the platform data. The parameters can then + be set through that platform data. An example of doing that is: + +#include <linux/ramoops.h> +[...] + +static struct ramoops_platform_data ramoops_data = { + .mem_size = <...>, + .mem_address = <...>, + .record_size = <...>, + .dump_oops = <...>, +}; + +static struct platform_device ramoops_dev = { + .name = "ramoops", + .dev = { + .platform_data = &ramoops_data, + }, +}; + +[... inside a function ...] +int ret; + +ret = platform_device_register(&ramoops_dev); +if (ret) { + printk(KERN_ERR "unable to register platform device\n"); + return ret; +} + +3. Dump format + +The data dump begins with a header, currently defined as "====" followed by a +timestamp and a new line. The dump then continues with the actual data. + +4. Reading the data + +The dump data can be read from memory (through /dev/mem or other means). +Getting the module parameters, which are needed in order to parse the data, can +be done through /sys/module/ramoops/parameters/* . diff --git a/Documentation/virtual/00-INDEX b/Documentation/virtual/00-INDEX index fe0251c4cfb7..8e601991d91c 100644 --- a/Documentation/virtual/00-INDEX +++ b/Documentation/virtual/00-INDEX @@ -8,3 +8,6 @@ lguest/ - Extremely simple hypervisor for experimental/educational use. uml/ - User Mode Linux, builds/runs Linux kernel as a userspace program. +virtio.txt + - Text version of draft virtio spec. + See http://ozlabs.org/~rusty/virtio-spec diff --git a/Documentation/virtual/lguest/lguest.c b/Documentation/virtual/lguest/lguest.c index 043bd7df3139..d928c134dee6 100644 --- a/Documentation/virtual/lguest/lguest.c +++ b/Documentation/virtual/lguest/lguest.c @@ -1996,6 +1996,9 @@ int main(int argc, char *argv[]) /* We use a simple helper to copy the arguments separated by spaces. */ concat((char *)(boot + 1), argv+optind+2); + /* Set kernel alignment to 16M (CONFIG_PHYSICAL_ALIGN) */ + boot->hdr.kernel_alignment = 0x1000000; + /* Boot protocol version: 2.07 supports the fields for lguest. */ boot->hdr.version = 0x207; diff --git a/Documentation/virtual/virtio-spec.txt b/Documentation/virtual/virtio-spec.txt new file mode 100644 index 000000000000..a350ae135b8c --- /dev/null +++ b/Documentation/virtual/virtio-spec.txt @@ -0,0 +1,2200 @@ +[Generated file: see http://ozlabs.org/~rusty/virtio-spec/] +Virtio PCI Card Specification +v0.9.1 DRAFT +- + +Rusty Russell <rusty@rustcorp.com.au>IBM Corporation (Editor) + +2011 August 1. + +Purpose and Description + +This document describes the specifications of the “virtio” family +of PCI[LaTeX Command: nomenclature] devices. These are devices +are found in virtual environments[LaTeX Command: nomenclature], +yet by design they are not all that different from physical PCI +devices, and this document treats them as such. This allows the +guest to use standard PCI drivers and discovery mechanisms. + +The purpose of virtio and this specification is that virtual +environments and guests should have a straightforward, efficient, +standard and extensible mechanism for virtual devices, rather +than boutique per-environment or per-OS mechanisms. + + Straightforward: Virtio PCI devices use normal PCI mechanisms + of interrupts and DMA which should be familiar to any device + driver author. There is no exotic page-flipping or COW + mechanism: it's just a PCI device.[footnote: +This lack of page-sharing implies that the implementation of the +device (e.g. the hypervisor or host) needs full access to the +guest memory. Communication with untrusted parties (i.e. +inter-guest communication) requires copying. +] + + Efficient: Virtio PCI devices consist of rings of descriptors + for input and output, which are neatly separated to avoid cache + effects from both guest and device writing to the same cache + lines. + + Standard: Virtio PCI makes no assumptions about the environment + in which it operates, beyond supporting PCI. In fact the virtio + devices specified in the appendices do not require PCI at all: + they have been implemented on non-PCI buses.[footnote: +The Linux implementation further separates the PCI virtio code +from the specific virtio drivers: these drivers are shared with +the non-PCI implementations (currently lguest and S/390). +] + + Extensible: Virtio PCI devices contain feature bits which are + acknowledged by the guest operating system during device setup. + This allows forwards and backwards compatibility: the device + offers all the features it knows about, and the driver + acknowledges those it understands and wishes to use. + + Virtqueues + +The mechanism for bulk data transport on virtio PCI devices is +pretentiously called a virtqueue. Each device can have zero or +more virtqueues: for example, the network device has one for +transmit and one for receive. + +Each virtqueue occupies two or more physically-contiguous pages +(defined, for the purposes of this specification, as 4096 bytes), +and consists of three parts: + + ++-------------------+-----------------------------------+-----------+ +| Descriptor Table | Available Ring (padding) | Used Ring | ++-------------------+-----------------------------------+-----------+ + + +When the driver wants to send buffers to the device, it puts them +in one or more slots in the descriptor table, and writes the +descriptor indices into the available ring. It then notifies the +device. When the device has finished with the buffers, it writes +the descriptors into the used ring, and sends an interrupt. + +Specification + + PCI Discovery + +Any PCI device with Vendor ID 0x1AF4, and Device ID 0x1000 +through 0x103F inclusive is a virtio device[footnote: +The actual value within this range is ignored +]. The device must also have a Revision ID of 0 to match this +specification. + +The Subsystem Device ID indicates which virtio device is +supported by the device. The Subsystem Vendor ID should reflect +the PCI Vendor ID of the environment (it's currently only used +for informational purposes by the guest). + + ++----------------------+--------------------+---------------+ +| Subsystem Device ID | Virtio Device | Specification | ++----------------------+--------------------+---------------+ ++----------------------+--------------------+---------------+ +| 1 | network card | Appendix C | ++----------------------+--------------------+---------------+ +| 2 | block device | Appendix D | ++----------------------+--------------------+---------------+ +| 3 | console | Appendix E | ++----------------------+--------------------+---------------+ +| 4 | entropy source | Appendix F | ++----------------------+--------------------+---------------+ +| 5 | memory ballooning | Appendix G | ++----------------------+--------------------+---------------+ +| 6 | ioMemory | - | ++----------------------+--------------------+---------------+ +| 9 | 9P transport | - | ++----------------------+--------------------+---------------+ + + + Device Configuration + +To configure the device, we use the first I/O region of the PCI +device. This contains a virtio header followed by a +device-specific region. + +There may be different widths of accesses to the I/O region; the “ +natural” access method for each field in the virtio header must +be used (i.e. 32-bit accesses for 32-bit fields, etc), but the +device-specific region can be accessed using any width accesses, +and should obtain the same results. + +Note that this is possible because while the virtio header is PCI +(i.e. little) endian, the device-specific region is encoded in +the native endian of the guest (where such distinction is +applicable). + + Device Initialization Sequence + +We start with an overview of device initialization, then expand +on the details of the device and how each step is preformed. + + Reset the device. This is not required on initial start up. + + The ACKNOWLEDGE status bit is set: we have noticed the device. + + The DRIVER status bit is set: we know how to drive the device. + + Device-specific setup, including reading the Device Feature + Bits, discovery of virtqueues for the device, optional MSI-X + setup, and reading and possibly writing the virtio + configuration space. + + The subset of Device Feature Bits understood by the driver is + written to the device. + + The DRIVER_OK status bit is set. + + The device can now be used (ie. buffers added to the + virtqueues)[footnote: +Historically, drivers have used the device before steps 5 and 6. +This is only allowed if the driver does not use any features +which would alter this early use of the device. +] + +If any of these steps go irrecoverably wrong, the guest should +set the FAILED status bit to indicate that it has given up on the +device (it can reset the device later to restart if desired). + +We now cover the fields required for general setup in detail. + + Virtio Header + +The virtio header looks as follows: + + ++------------++---------------------+---------------------+----------+--------+---------+---------+---------+--------+ +| Bits || 32 | 32 | 32 | 16 | 16 | 16 | 8 | 8 | ++------------++---------------------+---------------------+----------+--------+---------+---------+---------+--------+ +| Read/Write || R | R+W | R+W | R | R+W | R+W | R+W | R | ++------------++---------------------+---------------------+----------+--------+---------+---------+---------+--------+ +| Purpose || Device | Guest | Queue | Queue | Queue | Queue | Device | ISR | +| || Features bits 0:31 | Features bits 0:31 | Address | Size | Select | Notify | Status | Status | ++------------++---------------------+---------------------+----------+--------+---------+---------+---------+--------+ + + +If MSI-X is enabled for the device, two additional fields +immediately follow this header: + + ++------------++----------------+--------+ +| Bits || 16 | 16 | + +----------------+--------+ ++------------++----------------+--------+ +| Read/Write || R+W | R+W | ++------------++----------------+--------+ +| Purpose || Configuration | Queue | +| (MSI-X) || Vector | Vector | ++------------++----------------+--------+ + + +Finally, if feature bits (VIRTIO_F_FEATURES_HI) this is +immediately followed by two additional fields: + + ++------------++----------------------+---------------------- +| Bits || 32 | 32 ++------------++----------------------+---------------------- +| Read/Write || R | R+W ++------------++----------------------+---------------------- +| Purpose || Device | Guest +| || Features bits 32:63 | Features bits 32:63 ++------------++----------------------+---------------------- + + +Immediately following these general headers, there may be +device-specific headers: + + ++------------++--------------------+ +| Bits || Device Specific | + +--------------------+ ++------------++--------------------+ +| Read/Write || Device Specific | ++------------++--------------------+ +| Purpose || Device Specific... | +| || | ++------------++--------------------+ + + + Device Status + +The Device Status field is updated by the guest to indicate its +progress. This provides a simple low-level diagnostic: it's most +useful to imagine them hooked up to traffic lights on the console +indicating the status of each device. + +The device can be reset by writing a 0 to this field, otherwise +at least one bit should be set: + + ACKNOWLEDGE (1) Indicates that the guest OS has found the + device and recognized it as a valid virtio device. + + DRIVER (2) Indicates that the guest OS knows how to drive the + device. Under Linux, drivers can be loadable modules so there + may be a significant (or infinite) delay before setting this + bit. + + DRIVER_OK (3) Indicates that the driver is set up and ready to + drive the device. + + FAILED (8) Indicates that something went wrong in the guest, + and it has given up on the device. This could be an internal + error, or the driver didn't like the device for some reason, or + even a fatal error during device operation. The device must be + reset before attempting to re-initialize. + + Feature Bits + +The least significant 31 bits of the first configuration field +indicates the features that the device supports (the high bit is +reserved, and will be used to indicate the presence of future +feature bits elsewhere). If more than 31 feature bits are +supported, the device indicates so by setting feature bit 31 (see +[cha:Reserved-Feature-Bits]). The bits are allocated as follows: + + 0 to 23 Feature bits for the specific device type + + 24 to 40 Feature bits reserved for extensions to the queue and + feature negotiation mechanisms + + 41 to 63 Feature bits reserved for future extensions + +For example, feature bit 0 for a network device (i.e. Subsystem +Device ID 1) indicates that the device supports checksumming of +packets. + +The feature bits are negotiated: the device lists all the +features it understands in the Device Features field, and the +guest writes the subset that it understands into the Guest +Features field. The only way to renegotiate is to reset the +device. + +In particular, new fields in the device configuration header are +indicated by offering a feature bit, so the guest can check +before accessing that part of the configuration space. + +This allows for forwards and backwards compatibility: if the +device is enhanced with a new feature bit, older guests will not +write that feature bit back to the Guest Features field and it +can go into backwards compatibility mode. Similarly, if a guest +is enhanced with a feature that the device doesn't support, it +will not see that feature bit in the Device Features field and +can go into backwards compatibility mode (or, for poor +implementations, set the FAILED Device Status bit). + +Access to feature bits 32 to 63 is enabled by Guest by setting +feature bit 31. If this bit is unset, Device must assume that all +feature bits > 31 are unset. + + Configuration/Queue Vectors + +When MSI-X capability is present and enabled in the device +(through standard PCI configuration space) 4 bytes at byte offset +20 are used to map configuration change and queue interrupts to +MSI-X vectors. In this case, the ISR Status field is unused, and +device specific configuration starts at byte offset 24 in virtio +header structure. When MSI-X capability is not enabled, device +specific configuration starts at byte offset 20 in virtio header. + +Writing a valid MSI-X Table entry number, 0 to 0x7FF, to one of +Configuration/Queue Vector registers, maps interrupts triggered +by the configuration change/selected queue events respectively to +the corresponding MSI-X vector. To disable interrupts for a +specific event type, unmap it by writing a special NO_VECTOR +value: + +/* Vector value used to disable MSI for queue */ + +#define VIRTIO_MSI_NO_VECTOR 0xffff + +Reading these registers returns vector mapped to a given event, +or NO_VECTOR if unmapped. All queue and configuration change +events are unmapped by default. + +Note that mapping an event to vector might require allocating +internal device resources, and might fail. Devices report such +failures by returning the NO_VECTOR value when the relevant +Vector field is read. After mapping an event to vector, the +driver must verify success by reading the Vector field value: on +success, the previously written value is returned, and on +failure, NO_VECTOR is returned. If a mapping failure is detected, +the driver can retry mapping with fewervectors, or disable MSI-X. + + Virtqueue Configuration + +As a device can have zero or more virtqueues for bulk data +transport (for example, the network driver has two), the driver +needs to configure them as part of the device-specific +configuration. + +This is done as follows, for each virtqueue a device has: + + Write the virtqueue index (first queue is 0) to the Queue + Select field. + + Read the virtqueue size from the Queue Size field, which is + always a power of 2. This controls how big the virtqueue is + (see below). If this field is 0, the virtqueue does not exist. + + Allocate and zero virtqueue in contiguous physical memory, on a + 4096 byte alignment. Write the physical address, divided by + 4096 to the Queue Address field.[footnote: +The 4096 is based on the x86 page size, but it's also large +enough to ensure that the separate parts of the virtqueue are on +separate cache lines. +] + + Optionally, if MSI-X capability is present and enabled on the + device, select a vector to use to request interrupts triggered + by virtqueue events. Write the MSI-X Table entry number + corresponding to this vector in Queue Vector field. Read the + Queue Vector field: on success, previously written value is + returned; on failure, NO_VECTOR value is returned. + +The Queue Size field controls the total number of bytes required +for the virtqueue according to the following formula: + +#define ALIGN(x) (((x) + 4095) & ~4095) + +static inline unsigned vring_size(unsigned int qsz) + +{ + + return ALIGN(sizeof(struct vring_desc)*qsz + sizeof(u16)*(2 ++ qsz)) + + + ALIGN(sizeof(struct vring_used_elem)*qsz); + +} + +This currently wastes some space with padding, but also allows +future extensions. The virtqueue layout structure looks like this +(qsz is the Queue Size field, which is a variable, so this code +won't compile): + +struct vring { + + /* The actual descriptors (16 bytes each) */ + + struct vring_desc desc[qsz]; + + + + /* A ring of available descriptor heads with free-running +index. */ + + struct vring_avail avail; + + + + // Padding to the next 4096 boundary. + + char pad[]; + + + + // A ring of used descriptor heads with free-running index. + + struct vring_used used; + +}; + + A Note on Virtqueue Endianness + +Note that the endian of these fields and everything else in the +virtqueue is the native endian of the guest, not little-endian as +PCI normally is. This makes for simpler guest code, and it is +assumed that the host already has to be deeply aware of the guest +endian so such an “endian-aware” device is not a significant +issue. + + Descriptor Table + +The descriptor table refers to the buffers the guest is using for +the device. The addresses are physical addresses, and the buffers +can be chained via the next field. Each descriptor describes a +buffer which is read-only or write-only, but a chain of +descriptors can contain both read-only and write-only buffers. + +No descriptor chain may be more than 2^32 bytes long in total.struct vring_desc { + + /* Address (guest-physical). */ + + u64 addr; + + /* Length. */ + + u32 len; + +/* This marks a buffer as continuing via the next field. */ + +#define VRING_DESC_F_NEXT 1 + +/* This marks a buffer as write-only (otherwise read-only). */ + +#define VRING_DESC_F_WRITE 2 + +/* This means the buffer contains a list of buffer descriptors. +*/ + +#define VRING_DESC_F_INDIRECT 4 + + /* The flags as indicated above. */ + + u16 flags; + + /* Next field if flags & NEXT */ + + u16 next; + +}; + +The number of descriptors in the table is specified by the Queue +Size field for this virtqueue. + + <sub:Indirect-Descriptors>Indirect Descriptors + +Some devices benefit by concurrently dispatching a large number +of large requests. The VIRTIO_RING_F_INDIRECT_DESC feature can be +used to allow this (see [cha:Reserved-Feature-Bits]). To increase +ring capacity it is possible to store a table of indirect +descriptors anywhere in memory, and insert a descriptor in main +virtqueue (with flags&INDIRECT on) that refers to memory buffer +containing this indirect descriptor table; fields addr and len +refer to the indirect table address and length in bytes, +respectively. The indirect table layout structure looks like this +(len is the length of the descriptor that refers to this table, +which is a variable, so this code won't compile): + +struct indirect_descriptor_table { + + /* The actual descriptors (16 bytes each) */ + + struct vring_desc desc[len / 16]; + +}; + +The first indirect descriptor is located at start of the indirect +descriptor table (index 0), additional indirect descriptors are +chained by next field. An indirect descriptor without next field +(with flags&NEXT off) signals the end of the indirect descriptor +table, and transfers control back to the main virtqueue. An +indirect descriptor can not refer to another indirect descriptor +table (flags&INDIRECT must be off). A single indirect descriptor +table can include both read-only and write-only descriptors; +write-only flag (flags&WRITE) in the descriptor that refers to it +is ignored. + + Available Ring + +The available ring refers to what descriptors we are offering the +device: it refers to the head of a descriptor chain. The “flags” +field is currently 0 or 1: 1 indicating that we do not need an +interrupt when the device consumes a descriptor from the +available ring. Alternatively, the guest can ask the device to +delay interrupts until an entry with an index specified by the “ +used_event” field is written in the used ring (equivalently, +until the idx field in the used ring will reach the value +used_event + 1). The method employed by the device is controlled +by the VIRTIO_RING_F_EVENT_IDX feature bit (see [cha:Reserved-Feature-Bits] +). This interrupt suppression is merely an optimization; it may +not suppress interrupts entirely. + +The “idx” field indicates where we would put the next descriptor +entry (modulo the ring size). This starts at 0, and increases. + +struct vring_avail { + +#define VRING_AVAIL_F_NO_INTERRUPT 1 + + u16 flags; + + u16 idx; + + u16 ring[qsz]; /* qsz is the Queue Size field read from device +*/ + + u16 used_event; + +}; + + Used Ring + +The used ring is where the device returns buffers once it is done +with them. The flags field can be used by the device to hint that +no notification is necessary when the guest adds to the available +ring. Alternatively, the “avail_event” field can be used by the +device to hint that no notification is necessary until an entry +with an index specified by the “avail_event” is written in the +available ring (equivalently, until the idx field in the +available ring will reach the value avail_event + 1). The method +employed by the device is controlled by the guest through the +VIRTIO_RING_F_EVENT_IDX feature bit (see [cha:Reserved-Feature-Bits] +). [footnote: +These fields are kept here because this is the only part of the +virtqueue written by the device +]. + +Each entry in the ring is a pair: the head entry of the +descriptor chain describing the buffer (this matches an entry +placed in the available ring by the guest earlier), and the total +of bytes written into the buffer. The latter is extremely useful +for guests using untrusted buffers: if you do not know exactly +how much has been written by the device, you usually have to zero +the buffer to ensure no data leakage occurs. + +/* u32 is used here for ids for padding reasons. */ + +struct vring_used_elem { + + /* Index of start of used descriptor chain. */ + + u32 id; + + /* Total length of the descriptor chain which was used +(written to) */ + + u32 len; + +}; + + + +struct vring_used { + +#define VRING_USED_F_NO_NOTIFY 1 + + u16 flags; + + u16 idx; + + struct vring_used_elem ring[qsz]; + + u16 avail_event; + +}; + + Helpers for Managing Virtqueues + +The Linux Kernel Source code contains the definitions above and +helper routines in a more usable form, in +include/linux/virtio_ring.h. This was explicitly licensed by IBM +and Red Hat under the (3-clause) BSD license so that it can be +freely used by all other projects, and is reproduced (with slight +variation to remove Linux assumptions) in Appendix A. + + Device Operation + +There are two parts to device operation: supplying new buffers to +the device, and processing used buffers from the device. As an +example, the virtio network device has two virtqueues: the +transmit virtqueue and the receive virtqueue. The driver adds +outgoing (read-only) packets to the transmit virtqueue, and then +frees them after they are used. Similarly, incoming (write-only) +buffers are added to the receive virtqueue, and processed after +they are used. + + Supplying Buffers to The Device + +Actual transfer of buffers from the guest OS to the device +operates as follows: + + Place the buffer(s) into free descriptor(s). + + If there are no free descriptors, the guest may choose to + notify the device even if notifications are suppressed (to + reduce latency).[footnote: +The Linux drivers do this only for read-only buffers: for +write-only buffers, it is assumed that the driver is merely +trying to keep the receive buffer ring full, and no notification +of this expected condition is necessary. +] + + Place the id of the buffer in the next ring entry of the + available ring. + + The steps (1) and (2) may be performed repeatedly if batching + is possible. + + A memory barrier should be executed to ensure the device sees + the updated descriptor table and available ring before the next + step. + + The available “idx” field should be increased by the number of + entries added to the available ring. + + A memory barrier should be executed to ensure that we update + the idx field before checking for notification suppression. + + If notifications are not suppressed, the device should be + notified of the new buffers. + +Note that the above code does not take precautions against the +available ring buffer wrapping around: this is not possible since +the ring buffer is the same size as the descriptor table, so step +(1) will prevent such a condition. + +In addition, the maximum queue size is 32768 (it must be a power +of 2 which fits in 16 bits), so the 16-bit “idx” value can always +distinguish between a full and empty buffer. + +Here is a description of each stage in more detail. + + Placing Buffers Into The Descriptor Table + +A buffer consists of zero or more read-only physically-contiguous +elements followed by zero or more physically-contiguous +write-only elements (it must have at least one element). This +algorithm maps it into the descriptor table: + + for each buffer element, b: + + Get the next free descriptor table entry, d + + Set d.addr to the physical address of the start of b + + Set d.len to the length of b. + + If b is write-only, set d.flags to VRING_DESC_F_WRITE, + otherwise 0. + + If there is a buffer element after this: + + Set d.next to the index of the next free descriptor element. + + Set the VRING_DESC_F_NEXT bit in d.flags. + +In practice, the d.next fields are usually used to chain free +descriptors, and a separate count kept to check there are enough +free descriptors before beginning the mappings. + + Updating The Available Ring + +The head of the buffer we mapped is the first d in the algorithm +above. A naive implementation would do the following: + +avail->ring[avail->idx % qsz] = head; + +However, in general we can add many descriptors before we update +the “idx” field (at which point they become visible to the +device), so we keep a counter of how many we've added: + +avail->ring[(avail->idx + added++) % qsz] = head; + + Updating The Index Field + +Once the idx field of the virtqueue is updated, the device will +be able to access the descriptor entries we've created and the +memory they refer to. This is why a memory barrier is generally +used before the idx update, to ensure it sees the most up-to-date +copy. + +The idx field always increments, and we let it wrap naturally at +65536: + +avail->idx += added; + + <sub:Notifying-The-Device>Notifying The Device + +Device notification occurs by writing the 16-bit virtqueue index +of this virtqueue to the Queue Notify field of the virtio header +in the first I/O region of the PCI device. This can be expensive, +however, so the device can suppress such notifications if it +doesn't need them. We have to be careful to expose the new idx +value before checking the suppression flag: it's OK to notify +gratuitously, but not to omit a required notification. So again, +we use a memory barrier here before reading the flags or the +avail_event field. + +If the VIRTIO_F_RING_EVENT_IDX feature is not negotiated, and if +the VRING_USED_F_NOTIFY flag is not set, we go ahead and write to +the PCI configuration space. + +If the VIRTIO_F_RING_EVENT_IDX feature is negotiated, we read the +avail_event field in the available ring structure. If the +available index crossed_the avail_event field value since the +last notification, we go ahead and write to the PCI configuration +space. The avail_event field wraps naturally at 65536 as well: + +(u16)(new_idx - avail_event - 1) < (u16)(new_idx - old_idx) + + <sub:Receiving-Used-Buffers>Receiving Used Buffers From The + Device + +Once the device has used a buffer (read from or written to it, or +parts of both, depending on the nature of the virtqueue and the +device), it sends an interrupt, following an algorithm very +similar to the algorithm used for the driver to send the device a +buffer: + + Write the head descriptor number to the next field in the used + ring. + + Update the used ring idx. + + Determine whether an interrupt is necessary: + + If the VIRTIO_F_RING_EVENT_IDX feature is not negotiated: check + if f the VRING_AVAIL_F_NO_INTERRUPT flag is not set in avail- + >flags + + If the VIRTIO_F_RING_EVENT_IDX feature is negotiated: check + whether the used index crossed the used_event field value + since the last update. The used_event field wraps naturally + at 65536 as well:(u16)(new_idx - used_event - 1) < (u16)(new_idx - old_idx) + + If an interrupt is necessary: + + If MSI-X capability is disabled: + + Set the lower bit of the ISR Status field for the device. + + Send the appropriate PCI interrupt for the device. + + If MSI-X capability is enabled: + + Request the appropriate MSI-X interrupt message for the + device, Queue Vector field sets the MSI-X Table entry + number. + + If Queue Vector field value is NO_VECTOR, no interrupt + message is requested for this event. + +The guest interrupt handler should: + + If MSI-X capability is disabled: read the ISR Status field, + which will reset it to zero. If the lower bit is zero, the + interrupt was not for this device. Otherwise, the guest driver + should look through the used rings of each virtqueue for the + device, to see if any progress has been made by the device + which requires servicing. + + If MSI-X capability is enabled: look through the used rings of + each virtqueue mapped to the specific MSI-X vector for the + device, to see if any progress has been made by the device + which requires servicing. + +For each ring, guest should then disable interrupts by writing +VRING_AVAIL_F_NO_INTERRUPT flag in avail structure, if required. +It can then process used ring entries finally enabling interrupts +by clearing the VRING_AVAIL_F_NO_INTERRUPT flag or updating the +EVENT_IDX field in the available structure, Guest should then +execute a memory barrier, and then recheck the ring empty +condition. This is necessary to handle the case where, after the +last check and before enabling interrupts, an interrupt has been +suppressed by the device: + +vring_disable_interrupts(vq); + +for (;;) { + + if (vq->last_seen_used != vring->used.idx) { + + vring_enable_interrupts(vq); + + mb(); + + if (vq->last_seen_used != vring->used.idx) + + break; + + } + + struct vring_used_elem *e = +vring.used->ring[vq->last_seen_used%vsz]; + + process_buffer(e); + + vq->last_seen_used++; + +} + + Dealing With Configuration Changes + +Some virtio PCI devices can change the device configuration +state, as reflected in the virtio header in the PCI configuration +space. In this case: + + If MSI-X capability is disabled: an interrupt is delivered and + the second highest bit is set in the ISR Status field to + indicate that the driver should re-examine the configuration + space.Note that a single interrupt can indicate both that one + or more virtqueue has been used and that the configuration + space has changed: even if the config bit is set, virtqueues + must be scanned. + + If MSI-X capability is enabled: an interrupt message is + requested. The Configuration Vector field sets the MSI-X Table + entry number to use. If Configuration Vector field value is + NO_VECTOR, no interrupt message is requested for this event. + +Creating New Device Types + +Various considerations are necessary when creating a new device +type: + + How Many Virtqueues? + +It is possible that a very simple device will operate entirely +through its configuration space, but most will need at least one +virtqueue in which it will place requests. A device with both +input and output (eg. console and network devices described here) +need two queues: one which the driver fills with buffers to +receive input, and one which the driver places buffers to +transmit output. + + What Configuration Space Layout? + +Configuration space is generally used for rarely-changing or +initialization-time parameters. But it is a limited resource, so +it might be better to use a virtqueue to update configuration +information (the network device does this for filtering, +otherwise the table in the config space could potentially be very +large). + +Note that this space is generally the guest's native endian, +rather than PCI's little-endian. + + What Device Number? + +Currently device numbers are assigned quite freely: a simple +request mail to the author of this document or the Linux +virtualization mailing list[footnote: + +https://lists.linux-foundation.org/mailman/listinfo/virtualization +] will be sufficient to secure a unique one. + +Meanwhile for experimental drivers, use 65535 and work backwards. + + How many MSI-X vectors? + +Using the optional MSI-X capability devices can speed up +interrupt processing by removing the need to read ISR Status +register by guest driver (which might be an expensive operation), +reducing interrupt sharing between devices and queues within the +device, and handling interrupts from multiple CPUs. However, some +systems impose a limit (which might be as low as 256) on the +total number of MSI-X vectors that can be allocated to all +devices. Devices and/or device drivers should take this into +account, limiting the number of vectors used unless the device is +expected to cause a high volume of interrupts. Devices can +control the number of vectors used by limiting the MSI-X Table +Size or not presenting MSI-X capability in PCI configuration +space. Drivers can control this by mapping events to as small +number of vectors as possible, or disabling MSI-X capability +altogether. + + Message Framing + +The descriptors used for a buffer should not effect the semantics +of the message, except for the total length of the buffer. For +example, a network buffer consists of a 10 byte header followed +by the network packet. Whether this is presented in the ring +descriptor chain as (say) a 10 byte buffer and a 1514 byte +buffer, or a single 1524 byte buffer, or even three buffers, +should have no effect. + +In particular, no implementation should use the descriptor +boundaries to determine the size of any header in a request.[footnote: +The current qemu device implementations mistakenly insist that +the first descriptor cover the header in these cases exactly, so +a cautious driver should arrange it so. +] + + Device Improvements + +Any change to configuration space, or new virtqueues, or +behavioural changes, should be indicated by negotiation of a new +feature bit. This establishes clarity[footnote: +Even if it does mean documenting design or implementation +mistakes! +] and avoids future expansion problems. + +Clusters of functionality which are always implemented together +can use a single bit, but if one feature makes sense without the +others they should not be gratuitously grouped together to +conserve feature bits. We can always extend the spec when the +first person needs more than 24 feature bits for their device. + +[LaTeX Command: printnomenclature] + +Appendix A: virtio_ring.h + +#ifndef VIRTIO_RING_H + +#define VIRTIO_RING_H + +/* An interface for efficient virtio implementation. + + * + + * This header is BSD licensed so anyone can use the definitions + + * to implement compatible drivers/servers. + + * + + * Copyright 2007, 2009, IBM Corporation + + * Copyright 2011, Red Hat, Inc + + * All rights reserved. + + * + + * Redistribution and use in source and binary forms, with or +without + + * modification, are permitted provided that the following +conditions + + * are met: + + * 1. Redistributions of source code must retain the above +copyright + + * notice, this list of conditions and the following +disclaimer. + + * 2. 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. + + * 3. Neither the name of IBM nor the names of its contributors + + * may be used to endorse or promote products derived from +this software + + * without specific prior written permission. + + * 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 IBM 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. + + */ + + + +/* This marks a buffer as continuing via the next field. */ + +#define VRING_DESC_F_NEXT 1 + +/* This marks a buffer as write-only (otherwise read-only). */ + +#define VRING_DESC_F_WRITE 2 + + + +/* The Host uses this in used->flags to advise the Guest: don't +kick me + + * when you add a buffer. It's unreliable, so it's simply an + + * optimization. Guest will still kick if it's out of buffers. +*/ + +#define VRING_USED_F_NO_NOTIFY 1 + +/* The Guest uses this in avail->flags to advise the Host: don't + + * interrupt me when you consume a buffer. It's unreliable, so +it's + + * simply an optimization. */ + +#define VRING_AVAIL_F_NO_INTERRUPT 1 + + + +/* Virtio ring descriptors: 16 bytes. + + * These can chain together via "next". */ + +struct vring_desc { + + /* Address (guest-physical). */ + + uint64_t addr; + + /* Length. */ + + uint32_t len; + + /* The flags as indicated above. */ + + uint16_t flags; + + /* We chain unused descriptors via this, too */ + + uint16_t next; + +}; + + + +struct vring_avail { + + uint16_t flags; + + uint16_t idx; + + uint16_t ring[]; + + uint16_t used_event; + +}; + + + +/* u32 is used here for ids for padding reasons. */ + +struct vring_used_elem { + + /* Index of start of used descriptor chain. */ + + uint32_t id; + + /* Total length of the descriptor chain which was written +to. */ + + uint32_t len; + +}; + + + +struct vring_used { + + uint16_t flags; + + uint16_t idx; + + struct vring_used_elem ring[]; + + uint16_t avail_event; + +}; + + + +struct vring { + + unsigned int num; + + + + struct vring_desc *desc; + + struct vring_avail *avail; + + struct vring_used *used; + +}; + + + +/* The standard layout for the ring is a continuous chunk of +memory which + + * looks like this. We assume num is a power of 2. + + * + + * struct vring { + + * // The actual descriptors (16 bytes each) + + * struct vring_desc desc[num]; + + * + + * // A ring of available descriptor heads with free-running +index. + + * __u16 avail_flags; + + * __u16 avail_idx; + + * __u16 available[num]; + + * + + * // Padding to the next align boundary. + + * char pad[]; + + * + + * // A ring of used descriptor heads with free-running +index. + + * __u16 used_flags; + + * __u16 EVENT_IDX; + + * struct vring_used_elem used[num]; + + * }; + + * Note: for virtio PCI, align is 4096. + + */ + +static inline void vring_init(struct vring *vr, unsigned int num, +void *p, + + unsigned long align) + +{ + + vr->num = num; + + vr->desc = p; + + vr->avail = p + num*sizeof(struct vring_desc); + + vr->used = (void *)(((unsigned long)&vr->avail->ring[num] + + + align-1) + + & ~(align - 1)); + +} + + + +static inline unsigned vring_size(unsigned int num, unsigned long +align) + +{ + + return ((sizeof(struct vring_desc)*num + +sizeof(uint16_t)*(2+num) + + + align - 1) & ~(align - 1)) + + + sizeof(uint16_t)*3 + sizeof(struct +vring_used_elem)*num; + +} + + + +static inline int vring_need_event(uint16_t event_idx, uint16_t +new_idx, uint16_t old_idx) + +{ + + return (uint16_t)(new_idx - event_idx - 1) < +(uint16_t)(new_idx - old_idx); + +} + +#endif /* VIRTIO_RING_H */ + +<cha:Reserved-Feature-Bits>Appendix B: Reserved Feature Bits + +Currently there are five device-independent feature bits defined: + + VIRTIO_F_NOTIFY_ON_EMPTY (24) Negotiating this feature + indicates that the driver wants an interrupt if the device runs + out of available descriptors on a virtqueue, even though + interrupts are suppressed using the VRING_AVAIL_F_NO_INTERRUPT + flag or the used_event field. An example of this is the + networking driver: it doesn't need to know every time a packet + is transmitted, but it does need to free the transmitted + packets a finite time after they are transmitted. It can avoid + using a timer if the device interrupts it when all the packets + are transmitted. + + VIRTIO_F_RING_INDIRECT_DESC (28) Negotiating this feature + indicates that the driver can use descriptors with the + VRING_DESC_F_INDIRECT flag set, as described in [sub:Indirect-Descriptors] + . + + VIRTIO_F_RING_EVENT_IDX(29) This feature enables the used_event + and the avail_event fields. If set, it indicates that the + device should ignore the flags field in the available ring + structure. Instead, the used_event field in this structure is + used by guest to suppress device interrupts. Further, the + driver should ignore the flags field in the used ring + structure. Instead, the avail_event field in this structure is + used by the device to suppress notifications. If unset, the + driver should ignore the used_event field; the device should + ignore the avail_event field; the flags field is used + + VIRTIO_F_BAD_FEATURE(30) This feature should never be + negotiated by the guest; doing so is an indication that the + guest is faulty[footnote: +An experimental virtio PCI driver contained in Linux version +2.6.25 had this problem, and this feature bit can be used to +detect it. +] + + VIRTIO_F_FEATURES_HIGH(31) This feature indicates that the + device supports feature bits 32:63. If unset, feature bits + 32:63 are unset. + +Appendix C: Network Device + +The virtio network device is a virtual ethernet card, and is the +most complex of the devices supported so far by virtio. It has +enhanced rapidly and demonstrates clearly how support for new +features should be added to an existing device. Empty buffers are +placed in one virtqueue for receiving packets, and outgoing +packets are enqueued into another for transmission in that order. +A third command queue is used to control advanced filtering +features. + + Configuration + + Subsystem Device ID 1 + + Virtqueues 0:receiveq. 1:transmitq. 2:controlq[footnote: +Only if VIRTIO_NET_F_CTRL_VQ set +] + + Feature bits + + VIRTIO_NET_F_CSUM (0) Device handles packets with partial + checksum + + VIRTIO_NET_F_GUEST_CSUM (1) Guest handles packets with partial + checksum + + VIRTIO_NET_F_MAC (5) Device has given MAC address. + + VIRTIO_NET_F_GSO (6) (Deprecated) device handles packets with + any GSO type.[footnote: +It was supposed to indicate segmentation offload support, but +upon further investigation it became clear that multiple bits +were required. +] + + VIRTIO_NET_F_GUEST_TSO4 (7) Guest can receive TSOv4. + + VIRTIO_NET_F_GUEST_TSO6 (8) Guest can receive TSOv6. + + VIRTIO_NET_F_GUEST_ECN (9) Guest can receive TSO with ECN. + + VIRTIO_NET_F_GUEST_UFO (10) Guest can receive UFO. + + VIRTIO_NET_F_HOST_TSO4 (11) Device can receive TSOv4. + + VIRTIO_NET_F_HOST_TSO6 (12) Device can receive TSOv6. + + VIRTIO_NET_F_HOST_ECN (13) Device can receive TSO with ECN. + + VIRTIO_NET_F_HOST_UFO (14) Device can receive UFO. + + VIRTIO_NET_F_MRG_RXBUF (15) Guest can merge receive buffers. + + VIRTIO_NET_F_STATUS (16) Configuration status field is + available. + + VIRTIO_NET_F_CTRL_VQ (17) Control channel is available. + + VIRTIO_NET_F_CTRL_RX (18) Control channel RX mode support. + + VIRTIO_NET_F_CTRL_VLAN (19) Control channel VLAN filtering. + + Device configuration layout Two configuration fields are + currently defined. The mac address field always exists (though + is only valid if VIRTIO_NET_F_MAC is set), and the status field + only exists if VIRTIO_NET_F_STATUS is set. Only one bit is + currently defined for the status field: VIRTIO_NET_S_LINK_UP. #define VIRTIO_NET_S_LINK_UP 1 + + + +struct virtio_net_config { + + u8 mac[6]; + + u16 status; + +}; + + Device Initialization + + The initialization routine should identify the receive and + transmission virtqueues. + + If the VIRTIO_NET_F_MAC feature bit is set, the configuration + space “mac” entry indicates the “physical” address of the the + network card, otherwise a private MAC address should be + assigned. All guests are expected to negotiate this feature if + it is set. + + If the VIRTIO_NET_F_CTRL_VQ feature bit is negotiated, identify + the control virtqueue. + + If the VIRTIO_NET_F_STATUS feature bit is negotiated, the link + status can be read from the bottom bit of the “status” config + field. Otherwise, the link should be assumed active. + + The receive virtqueue should be filled with receive buffers. + This is described in detail below in “Setting Up Receive + Buffers”. + + A driver can indicate that it will generate checksumless + packets by negotating the VIRTIO_NET_F_CSUM feature. This “ + checksum offload” is a common feature on modern network cards. + + If that feature is negotiated, a driver can use TCP or UDP + segmentation offload by negotiating the VIRTIO_NET_F_HOST_TSO4 + (IPv4 TCP), VIRTIO_NET_F_HOST_TSO6 (IPv6 TCP) and + VIRTIO_NET_F_HOST_UFO (UDP fragmentation) features. It should + not send TCP packets requiring segmentation offload which have + the Explicit Congestion Notification bit set, unless the + VIRTIO_NET_F_HOST_ECN feature is negotiated.[footnote: +This is a common restriction in real, older network cards. +] + + The converse features are also available: a driver can save the + virtual device some work by negotiating these features.[footnote: +For example, a network packet transported between two guests on +the same system may not require checksumming at all, nor +segmentation, if both guests are amenable. +] The VIRTIO_NET_F_GUEST_CSUM feature indicates that partially + checksummed packets can be received, and if it can do that then + the VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6, + VIRTIO_NET_F_GUEST_UFO and VIRTIO_NET_F_GUEST_ECN are the input + equivalents of the features described above. See “Receiving + Packets” below. + + Device Operation + +Packets are transmitted by placing them in the transmitq, and +buffers for incoming packets are placed in the receiveq. In each +case, the packet itself is preceeded by a header: + +struct virtio_net_hdr { + +#define VIRTIO_NET_HDR_F_NEEDS_CSUM 1 + + u8 flags; + +#define VIRTIO_NET_HDR_GSO_NONE 0 + +#define VIRTIO_NET_HDR_GSO_TCPV4 1 + +#define VIRTIO_NET_HDR_GSO_UDP 3 + +#define VIRTIO_NET_HDR_GSO_TCPV6 4 + +#define VIRTIO_NET_HDR_GSO_ECN 0x80 + + u8 gso_type; + + u16 hdr_len; + + u16 gso_size; + + u16 csum_start; + + u16 csum_offset; + +/* Only if VIRTIO_NET_F_MRG_RXBUF: */ + + u16 num_buffers + +}; + +The controlq is used to control device features such as +filtering. + + Packet Transmission + +Transmitting a single packet is simple, but varies depending on +the different features the driver negotiated. + + If the driver negotiated VIRTIO_NET_F_CSUM, and the packet has + not been fully checksummed, then the virtio_net_hdr's fields + are set as follows. Otherwise, the packet must be fully + checksummed, and flags is zero. + + flags has the VIRTIO_NET_HDR_F_NEEDS_CSUM set, + + <ite:csum_start-is-set>csum_start is set to the offset within + the packet to begin checksumming, and + + csum_offset indicates how many bytes after the csum_start the + new (16 bit ones' complement) checksum should be placed.[footnote: +For example, consider a partially checksummed TCP (IPv4) packet. +It will have a 14 byte ethernet header and 20 byte IP header +followed by the TCP header (with the TCP checksum field 16 bytes +into that header). csum_start will be 14+20 = 34 (the TCP +checksum includes the header), and csum_offset will be 16. The +value in the TCP checksum field will be the sum of the TCP pseudo +header, so that replacing it by the ones' complement checksum of +the TCP header and body will give the correct result. +] + + <enu:If-the-driver>If the driver negotiated + VIRTIO_NET_F_HOST_TSO4, TSO6 or UFO, and the packet requires + TCP segmentation or UDP fragmentation, then the “gso_type” + field is set to VIRTIO_NET_HDR_GSO_TCPV4, TCPV6 or UDP. + (Otherwise, it is set to VIRTIO_NET_HDR_GSO_NONE). In this + case, packets larger than 1514 bytes can be transmitted: the + metadata indicates how to replicate the packet header to cut it + into smaller packets. The other gso fields are set: + + hdr_len is a hint to the device as to how much of the header + needs to be kept to copy into each packet, usually set to the + length of the headers, including the transport header.[footnote: +Due to various bugs in implementations, this field is not useful +as a guarantee of the transport header size. +] + + gso_size is the size of the packet beyond that header (ie. + MSS). + + If the driver negotiated the VIRTIO_NET_F_HOST_ECN feature, the + VIRTIO_NET_HDR_GSO_ECN bit may be set in “gso_type” as well, + indicating that the TCP packet has the ECN bit set.[footnote: +This case is not handled by some older hardware, so is called out +specifically in the protocol. +] + + If the driver negotiated the VIRTIO_NET_F_MRG_RXBUF feature, + the num_buffers field is set to zero. + + The header and packet are added as one output buffer to the + transmitq, and the device is notified of the new entry (see [sub:Notifying-The-Device] + ).[footnote: +Note that the header will be two bytes longer for the +VIRTIO_NET_F_MRG_RXBUF case. +] + + Packet Transmission Interrupt + +Often a driver will suppress transmission interrupts using the +VRING_AVAIL_F_NO_INTERRUPT flag (see [sub:Receiving-Used-Buffers] +) and check for used packets in the transmit path of following +packets. However, it will still receive interrupts if the +VIRTIO_F_NOTIFY_ON_EMPTY feature is negotiated, indicating that +the transmission queue is completely emptied. + +The normal behavior in this interrupt handler is to retrieve and +new descriptors from the used ring and free the corresponding +headers and packets. + + Setting Up Receive Buffers + +It is generally a good idea to keep the receive virtqueue as +fully populated as possible: if it runs out, network performance +will suffer. + +If the VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6 or +VIRTIO_NET_F_GUEST_UFO features are used, the Guest will need to +accept packets of up to 65550 bytes long (the maximum size of a +TCP or UDP packet, plus the 14 byte ethernet header), otherwise +1514 bytes. So unless VIRTIO_NET_F_MRG_RXBUF is negotiated, every +buffer in the receive queue needs to be at least this length [footnote: +Obviously each one can be split across multiple descriptor +elements. +]. + +If VIRTIO_NET_F_MRG_RXBUF is negotiated, each buffer must be at +least the size of the struct virtio_net_hdr. + + Packet Receive Interrupt + +When a packet is copied into a buffer in the receiveq, the +optimal path is to disable further interrupts for the receiveq +(see [sub:Receiving-Used-Buffers]) and process packets until no +more are found, then re-enable them. + +Processing packet involves: + + If the driver negotiated the VIRTIO_NET_F_MRG_RXBUF feature, + then the “num_buffers” field indicates how many descriptors + this packet is spread over (including this one). This allows + receipt of large packets without having to allocate large + buffers. In this case, there will be at least “num_buffers” in + the used ring, and they should be chained together to form a + single packet. The other buffers will not begin with a struct + virtio_net_hdr. + + If the VIRTIO_NET_F_MRG_RXBUF feature was not negotiated, or + the “num_buffers” field is one, then the entire packet will be + contained within this buffer, immediately following the struct + virtio_net_hdr. + + If the VIRTIO_NET_F_GUEST_CSUM feature was negotiated, the + VIRTIO_NET_HDR_F_NEEDS_CSUM bit in the “flags” field may be + set: if so, the checksum on the packet is incomplete and the “ + csum_start” and “csum_offset” fields indicate how to calculate + it (see [ite:csum_start-is-set]). + + If the VIRTIO_NET_F_GUEST_TSO4, TSO6 or UFO options were + negotiated, then the “gso_type” may be something other than + VIRTIO_NET_HDR_GSO_NONE, and the “gso_size” field indicates the + desired MSS (see [enu:If-the-driver]).Control Virtqueue + +The driver uses the control virtqueue (if VIRTIO_NET_F_VTRL_VQ is +negotiated) to send commands to manipulate various features of +the device which would not easily map into the configuration +space. + +All commands are of the following form: + +struct virtio_net_ctrl { + + u8 class; + + u8 command; + + u8 command-specific-data[]; + + u8 ack; + +}; + + + +/* ack values */ + +#define VIRTIO_NET_OK 0 + +#define VIRTIO_NET_ERR 1 + +The class, command and command-specific-data are set by the +driver, and the device sets the ack byte. There is little it can +do except issue a diagnostic if the ack byte is not +VIRTIO_NET_OK. + + Packet Receive Filtering + +If the VIRTIO_NET_F_CTRL_RX feature is negotiated, the driver can +send control commands for promiscuous mode, multicast receiving, +and filtering of MAC addresses. + +Note that in general, these commands are best-effort: unwanted +packets may still arrive. + + Setting Promiscuous Mode + +#define VIRTIO_NET_CTRL_RX 0 + + #define VIRTIO_NET_CTRL_RX_PROMISC 0 + + #define VIRTIO_NET_CTRL_RX_ALLMULTI 1 + +The class VIRTIO_NET_CTRL_RX has two commands: +VIRTIO_NET_CTRL_RX_PROMISC turns promiscuous mode on and off, and +VIRTIO_NET_CTRL_RX_ALLMULTI turns all-multicast receive on and +off. The command-specific-data is one byte containing 0 (off) or +1 (on). + + Setting MAC Address Filtering + +struct virtio_net_ctrl_mac { + + u32 entries; + + u8 macs[entries][ETH_ALEN]; + +}; + + + +#define VIRTIO_NET_CTRL_MAC 1 + + #define VIRTIO_NET_CTRL_MAC_TABLE_SET 0 + +The device can filter incoming packets by any number of +destination MAC addresses.[footnote: +Since there are no guarentees, it can use a hash filter +orsilently switch to allmulti or promiscuous mode if it is given +too many addresses. +] This table is set using the class VIRTIO_NET_CTRL_MAC and the +command VIRTIO_NET_CTRL_MAC_TABLE_SET. The command-specific-data +is two variable length tables of 6-byte MAC addresses. The first +table contains unicast addresses, and the second contains +multicast addresses. + + VLAN Filtering + +If the driver negotiates the VIRTION_NET_F_CTRL_VLAN feature, it +can control a VLAN filter table in the device. + +#define VIRTIO_NET_CTRL_VLAN 2 + + #define VIRTIO_NET_CTRL_VLAN_ADD 0 + + #define VIRTIO_NET_CTRL_VLAN_DEL 1 + +Both the VIRTIO_NET_CTRL_VLAN_ADD and VIRTIO_NET_CTRL_VLAN_DEL +command take a 16-bit VLAN id as the command-specific-data. + +Appendix D: Block Device + +The virtio block device is a simple virtual block device (ie. +disk). Read and write requests (and other exotic requests) are +placed in the queue, and serviced (probably out of order) by the +device except where noted. + + Configuration + + Subsystem Device ID 2 + + Virtqueues 0:requestq. + + Feature bits + + VIRTIO_BLK_F_BARRIER (0) Host supports request barriers. + + VIRTIO_BLK_F_SIZE_MAX (1) Maximum size of any single segment is + in “size_max”. + + VIRTIO_BLK_F_SEG_MAX (2) Maximum number of segments in a + request is in “seg_max”. + + VIRTIO_BLK_F_GEOMETRY (4) Disk-style geometry specified in “ + geometry”. + + VIRTIO_BLK_F_RO (5) Device is read-only. + + VIRTIO_BLK_F_BLK_SIZE (6) Block size of disk is in “blk_size”. + + VIRTIO_BLK_F_SCSI (7) Device supports scsi packet commands. + + VIRTIO_BLK_F_FLUSH (9) Cache flush command support. + + + + Device configuration layout The capacity of the device + (expressed in 512-byte sectors) is always present. The + availability of the others all depend on various feature bits + as indicated above. struct virtio_blk_config { + + u64 capacity; + + u32 size_max; + + u32 seg_max; + + struct virtio_blk_geometry { + + u16 cylinders; + + u8 heads; + + u8 sectors; + + } geometry; + + u32 blk_size; + + + +}; + + Device Initialization + + The device size should be read from the “capacity” + configuration field. No requests should be submitted which goes + beyond this limit. + + If the VIRTIO_BLK_F_BLK_SIZE feature is negotiated, the + blk_size field can be read to determine the optimal sector size + for the driver to use. This does not effect the units used in + the protocol (always 512 bytes), but awareness of the correct + value can effect performance. + + If the VIRTIO_BLK_F_RO feature is set by the device, any write + requests will fail. + + + + Device Operation + +The driver queues requests to the virtqueue, and they are used by +the device (not necessarily in order). Each request is of form: + +struct virtio_blk_req { + + + + u32 type; + + u32 ioprio; + + u64 sector; + + char data[][512]; + + u8 status; + +}; + +If the device has VIRTIO_BLK_F_SCSI feature, it can also support +scsi packet command requests, each of these requests is of form:struct virtio_scsi_pc_req { + + u32 type; + + u32 ioprio; + + u64 sector; + + char cmd[]; + + char data[][512]; + +#define SCSI_SENSE_BUFFERSIZE 96 + + u8 sense[SCSI_SENSE_BUFFERSIZE]; + + u32 errors; + + u32 data_len; + + u32 sense_len; + + u32 residual; + + u8 status; + +}; + +The type of the request is either a read (VIRTIO_BLK_T_IN), a +write (VIRTIO_BLK_T_OUT), a scsi packet command +(VIRTIO_BLK_T_SCSI_CMD or VIRTIO_BLK_T_SCSI_CMD_OUT[footnote: +the SCSI_CMD and SCSI_CMD_OUT types are equivalent, the device +does not distinguish between them +]) or a flush (VIRTIO_BLK_T_FLUSH or VIRTIO_BLK_T_FLUSH_OUT[footnote: +the FLUSH and FLUSH_OUT types are equivalent, the device does not +distinguish between them +]). If the device has VIRTIO_BLK_F_BARRIER feature the high bit +(VIRTIO_BLK_T_BARRIER) indicates that this request acts as a +barrier and that all preceeding requests must be complete before +this one, and all following requests must not be started until +this is complete. Note that a barrier does not flush caches in +the underlying backend device in host, and thus does not serve as +data consistency guarantee. Driver must use FLUSH request to +flush the host cache. + +#define VIRTIO_BLK_T_IN 0 + +#define VIRTIO_BLK_T_OUT 1 + +#define VIRTIO_BLK_T_SCSI_CMD 2 + +#define VIRTIO_BLK_T_SCSI_CMD_OUT 3 + +#define VIRTIO_BLK_T_FLUSH 4 + +#define VIRTIO_BLK_T_FLUSH_OUT 5 + +#define VIRTIO_BLK_T_BARRIER 0x80000000 + +The ioprio field is a hint about the relative priorities of +requests to the device: higher numbers indicate more important +requests. + +The sector number indicates the offset (multiplied by 512) where +the read or write is to occur. This field is unused and set to 0 +for scsi packet commands and for flush commands. + +The cmd field is only present for scsi packet command requests, +and indicates the command to perform. This field must reside in a +single, separate read-only buffer; command length can be derived +from the length of this buffer. + +Note that these first three (four for scsi packet commands) +fields are always read-only: the data field is either read-only +or write-only, depending on the request. The size of the read or +write can be derived from the total size of the request buffers. + +The sense field is only present for scsi packet command requests, +and indicates the buffer for scsi sense data. + +The data_len field is only present for scsi packet command +requests, this field is deprecated, and should be ignored by the +driver. Historically, devices copied data length there. + +The sense_len field is only present for scsi packet command +requests and indicates the number of bytes actually written to +the sense buffer. + +The residual field is only present for scsi packet command +requests and indicates the residual size, calculated as data +length - number of bytes actually transferred. + +The final status byte is written by the device: either +VIRTIO_BLK_S_OK for success, VIRTIO_BLK_S_IOERR for host or guest +error or VIRTIO_BLK_S_UNSUPP for a request unsupported by host:#define VIRTIO_BLK_S_OK 0 + +#define VIRTIO_BLK_S_IOERR 1 + +#define VIRTIO_BLK_S_UNSUPP 2 + +Historically, devices assumed that the fields type, ioprio and +sector reside in a single, separate read-only buffer; the fields +errors, data_len, sense_len and residual reside in a single, +separate write-only buffer; the sense field in a separate +write-only buffer of size 96 bytes, by itself; the fields errors, +data_len, sense_len and residual in a single write-only buffer; +and the status field is a separate read-only buffer of size 1 +byte, by itself. + +Appendix E: Console Device + +The virtio console device is a simple device for data input and +output. A device may have one or more ports. Each port has a pair +of input and output virtqueues. Moreover, a device has a pair of +control IO virtqueues. The control virtqueues are used to +communicate information between the device and the driver about +ports being opened and closed on either side of the connection, +indication from the host about whether a particular port is a +console port, adding new ports, port hot-plug/unplug, etc., and +indication from the guest about whether a port or a device was +successfully added, port open/close, etc.. For data IO, one or +more empty buffers are placed in the receive queue for incoming +data and outgoing characters are placed in the transmit queue. + + Configuration + + Subsystem Device ID 3 + + Virtqueues 0:receiveq(port0). 1:transmitq(port0), 2:control + receiveq[footnote: +Ports 2 onwards only if VIRTIO_CONSOLE_F_MULTIPORT is set +], 3:control transmitq, 4:receiveq(port1), 5:transmitq(port1), + ... + + Feature bits + + VIRTIO_CONSOLE_F_SIZE (0) Configuration cols and rows fields + are valid. + + VIRTIO_CONSOLE_F_MULTIPORT(1) Device has support for multiple + ports; configuration fields nr_ports and max_nr_ports are + valid and control virtqueues will be used. + + Device configuration layout The size of the console is supplied + in the configuration space if the VIRTIO_CONSOLE_F_SIZE feature + is set. Furthermore, if the VIRTIO_CONSOLE_F_MULTIPORT feature + is set, the maximum number of ports supported by the device can + be fetched.struct virtio_console_config { + + u16 cols; + + u16 rows; + + + + u32 max_nr_ports; + +}; + + Device Initialization + + If the VIRTIO_CONSOLE_F_SIZE feature is negotiated, the driver + can read the console dimensions from the configuration fields. + + If the VIRTIO_CONSOLE_F_MULTIPORT feature is negotiated, the + driver can spawn multiple ports, not all of which may be + attached to a console. Some could be generic ports. In this + case, the control virtqueues are enabled and according to the + max_nr_ports configuration-space value, the appropriate number + of virtqueues are created. A control message indicating the + driver is ready is sent to the host. The host can then send + control messages for adding new ports to the device. After + creating and initializing each port, a + VIRTIO_CONSOLE_PORT_READY control message is sent to the host + for that port so the host can let us know of any additional + configuration options set for that port. + + The receiveq for each port is populated with one or more + receive buffers. + + Device Operation + + For output, a buffer containing the characters is placed in the + port's transmitq.[footnote: +Because this is high importance and low bandwidth, the current +Linux implementation polls for the buffer to be used, rather than +waiting for an interrupt, simplifying the implementation +significantly. However, for generic serial ports with the +O_NONBLOCK flag set, the polling limitation is relaxed and the +consumed buffers are freed upon the next write or poll call or +when a port is closed or hot-unplugged. +] + + When a buffer is used in the receiveq (signalled by an + interrupt), the contents is the input to the port associated + with the virtqueue for which the notification was received. + + If the driver negotiated the VIRTIO_CONSOLE_F_SIZE feature, a + configuration change interrupt may occur. The updated size can + be read from the configuration fields. + + If the driver negotiated the VIRTIO_CONSOLE_F_MULTIPORT + feature, active ports are announced by the host using the + VIRTIO_CONSOLE_PORT_ADD control message. The same message is + used for port hot-plug as well. + + If the host specified a port `name', a sysfs attribute is + created with the name filled in, so that udev rules can be + written that can create a symlink from the port's name to the + char device for port discovery by applications in the guest. + + Changes to ports' state are effected by control messages. + Appropriate action is taken on the port indicated in the + control message. The layout of the structure of the control + buffer and the events associated are:struct virtio_console_control { + + uint32_t id; /* Port number */ + + uint16_t event; /* The kind of control event */ + + uint16_t value; /* Extra information for the event */ + +}; + + + +/* Some events for the internal messages (control packets) */ + + + +#define VIRTIO_CONSOLE_DEVICE_READY 0 + +#define VIRTIO_CONSOLE_PORT_ADD 1 + +#define VIRTIO_CONSOLE_PORT_REMOVE 2 + +#define VIRTIO_CONSOLE_PORT_READY 3 + +#define VIRTIO_CONSOLE_CONSOLE_PORT 4 + +#define VIRTIO_CONSOLE_RESIZE 5 + +#define VIRTIO_CONSOLE_PORT_OPEN 6 + +#define VIRTIO_CONSOLE_PORT_NAME 7 + +Appendix F: Entropy Device + +The virtio entropy device supplies high-quality randomness for +guest use. + + Configuration + + Subsystem Device ID 4 + + Virtqueues 0:requestq. + + Feature bits None currently defined + + Device configuration layout None currently defined. + + Device Initialization + + The virtqueue is initialized + + Device Operation + +When the driver requires random bytes, it places the descriptor +of one or more buffers in the queue. It will be completely filled +by random data by the device. + +Appendix G: Memory Balloon Device + +The virtio memory balloon device is a primitive device for +managing guest memory: the device asks for a certain amount of +memory, and the guest supplies it (or withdraws it, if the device +has more than it asks for). This allows the guest to adapt to +changes in allowance of underlying physical memory. If the +feature is negotiated, the device can also be used to communicate +guest memory statistics to the host. + + Configuration + + Subsystem Device ID 5 + + Virtqueues 0:inflateq. 1:deflateq. 2:statsq.[footnote: +Only if VIRTIO_BALLON_F_STATS_VQ set +] + + Feature bits + + VIRTIO_BALLOON_F_MUST_TELL_HOST (0) Host must be told before + pages from the balloon are used. + + VIRTIO_BALLOON_F_STATS_VQ (1) A virtqueue for reporting guest + memory statistics is present. + + Device configuration layout Both fields of this configuration + are always available. Note that they are little endian, despite + convention that device fields are guest endian:struct virtio_balloon_config { + + u32 num_pages; + + u32 actual; + +}; + + Device Initialization + + The inflate and deflate virtqueues are identified. + + If the VIRTIO_BALLOON_F_STATS_VQ feature bit is negotiated: + + Identify the stats virtqueue. + + Add one empty buffer to the stats virtqueue and notify the + host. + +Device operation begins immediately. + + Device Operation + + Memory Ballooning The device is driven by the receipt of a + configuration change interrupt. + + The “num_pages” configuration field is examined. If this is + greater than the “actual” number of pages, memory must be given + to the balloon. If it is less than the “actual” number of + pages, memory may be taken back from the balloon for general + use. + + To supply memory to the balloon (aka. inflate): + + The driver constructs an array of addresses of unused memory + pages. These addresses are divided by 4096[footnote: +This is historical, and independent of the guest page size +] and the descriptor describing the resulting 32-bit array is + added to the inflateq. + + To remove memory from the balloon (aka. deflate): + + The driver constructs an array of addresses of memory pages it + has previously given to the balloon, as described above. This + descriptor is added to the deflateq. + + If the VIRTIO_BALLOON_F_MUST_TELL_HOST feature is set, the + guest may not use these requested pages until that descriptor + in the deflateq has been used by the device. + + Otherwise, the guest may begin to re-use pages previously given + to the balloon before the device has acknowledged their + withdrawl. [footnote: +In this case, deflation advice is merely a courtesy +] + + In either case, once the device has completed the inflation or + deflation, the “actual” field of the configuration should be + updated to reflect the new number of pages in the balloon.[footnote: +As updates to configuration space are not atomic, this field +isn't particularly reliable, but can be used to diagnose buggy +guests. +] + + Memory Statistics + +The stats virtqueue is atypical because communication is driven +by the device (not the driver). The channel becomes active at +driver initialization time when the driver adds an empty buffer +and notifies the device. A request for memory statistics proceeds +as follows: + + The device pushes the buffer onto the used ring and sends an + interrupt. + + The driver pops the used buffer and discards it. + + The driver collects memory statistics and writes them into a + new buffer. + + The driver adds the buffer to the virtqueue and notifies the + device. + + The device pops the buffer (retaining it to initiate a + subsequent request) and consumes the statistics. + + Memory Statistics Format Each statistic consists of a 16 bit + tag and a 64 bit value. Both quantities are represented in the + native endian of the guest. All statistics are optional and the + driver may choose which ones to supply. To guarantee backwards + compatibility, unsupported statistics should be omitted. + + struct virtio_balloon_stat { + +#define VIRTIO_BALLOON_S_SWAP_IN 0 + +#define VIRTIO_BALLOON_S_SWAP_OUT 1 + +#define VIRTIO_BALLOON_S_MAJFLT 2 + +#define VIRTIO_BALLOON_S_MINFLT 3 + +#define VIRTIO_BALLOON_S_MEMFREE 4 + +#define VIRTIO_BALLOON_S_MEMTOT 5 + + u16 tag; + + u64 val; + +} __attribute__((packed)); + + Tags + + VIRTIO_BALLOON_S_SWAP_IN The amount of memory that has been + swapped in (in bytes). + + VIRTIO_BALLOON_S_SWAP_OUT The amount of memory that has been + swapped out to disk (in bytes). + + VIRTIO_BALLOON_S_MAJFLT The number of major page faults that + have occurred. + + VIRTIO_BALLOON_S_MINFLT The number of minor page faults that + have occurred. + + VIRTIO_BALLOON_S_MEMFREE The amount of memory not being used + for any purpose (in bytes). + + VIRTIO_BALLOON_S_MEMTOT The total amount of memory available + (in bytes). + |