test

Writing Device Drivers

Responsibilities of the Driver Writer

The developer must take responsibility for:

Device Driver Instances

The Solaris kernel allows multiple instances of a driver. Each instance has its own data space but shares the text and some global data with other instances. The device is managed on a per-instance basis and hardened drivers should use a separate instance for each piece of hardware, unless the driver is designed to handle fail-over internally. There can be multiple instances of a driver per slot--for example, multifunction cards--which is standard behavior for Solaris device drivers.

Exclusive Use of DDI Access Handles

All programmed I/O (PIO) access by a hardened driver must use DDI access functions from the ddi_get, ddi_put, ddi_rep_get, and ddi_rep_put families of routines. It should not directly access the mapped registers by the address returned from ddi_regs_map_setup(9F). Using an access handle ensures that an I/O fault is controlled and its effects confined to the returned value, rather than possibly corrupting other parts of the machine state. (Avoid the ddi_peek(9F) and ddi_poke(9F) routines because they do not use access handles.)

The DDI access mechanism is important because it provides an opportunity to control how data is read into the kernel. DDI access routines provide protection by constraining the effect of bus timeout traps.

Detecting Corrupted Data

The following subsections consider where data corruption can occur and the steps you can take to detect it.

Corruption of Device Management and Control Data

The driver should assume that any data obtained from the device, whether by PIO or DMA, could have been corrupted. In particular, extreme care should be taken with pointers, memory offsets, or array indexes read or calculated from data supplied by the device. Such values can be malignant, meaning they can cause a kernel panic if dereferenced. All such values should be checked for range and alignment (if required) before use.

Even if a pointer is not malignant, it can still be misleading. For example, it can point at a valid instance of an object, but not the correct one. Where possible, the driver should cross-check the pointer with the pointed-to object, or otherwise validate the data obtained through it.

Other types of data can also be misleading, such as packet lengths, status words, or channel IDs. Each should be checked to the extent possible: a packet length can be range-checked to ensure that it is not negative or larger than the containing buffer; a status word can be checked for "impossible" bits; and a channel ID can be matched against a list of valid IDs.

Where a value is used to identify a STREAM, the driver must ensure that the STREAM still exists. The asynchronous nature of STREAMS processing means that a STREAM can be dismantled while device interrupts are still outstanding.

The driver should not reread data from the device; the data should be read once, validated, and stored in the driver's local state. This avoids the hazard presented by data that, although correct when initally read and validated, is incorrect when reread later.

The driver should also ensure that all loops are bounded, so that a device returning a continuous BUSY status, or claiming that another buffer needs to be processed, does not lock up the entire system.

Corruption of Received Data

Device errors can result in corrupted data being placed in receive buffers. Such corruption is indistinguishable from corruption that occurs beyond the domain of the device--for example, within a network. Typically, existing software is already in place to handle such corruption, for example, through integrity checks at the transport layer of a protocol stack or within the application using the device.

If the received data will not be checked for integrity at a higher layer--as in the case of a disk driver, for example--it can be integrity checked within the driver itself. Methods of detecting corruption in received data are typically device-specific (checksums, CRC, and so forth).

Detecting Faults

Any ancestor of a device driver can disable the data path to the device if it detects a fault. When PIO access is disabled, reads from the device return undefined values, while writes are ignored. If DMA access is disabled, the device might be prevented from accessing memory, or it might receive undefined data on reads and have writes discarded.

For a device driver to detect that a data path has been disabled, the following DDI routines are provided:

Each function checks whether any faults affecting the data path, represented by the supplied handle, have been detected. If one of these functions returns DDI_FAILURE, indicating that the data path has failed, the driver should report the fault using ddi_dev_report_fault(9F), perform any necessary cleanup, and, where possible, return an appropriate error to its caller.

Containment of Faults

Preservation of system integrity requires that faults are detected before they alter the system state. Consequently, the driver must be sure to test for faults whenever data returned from the device is going to be used by the system.

Handling Stuck Interrupts

The driver must identify stuck interrupts because a persistently asserted interrupt severely affects system performance, almost certainly stalling a single-processor machine.

Sometimes the driver has a hard time identifying a particular interrupt as a hoax. For network drivers, if a receive interrupt is indicated but no new buffers have been made available, no work was needed. When this is an isolated occurrence, it is not a problem, as the actual work might already have been completed by another routine (read service, for example).

On the other hand, continuous interrupts with no work for the driver to process can indicate a stuck interrupt line. For this reason, all platforms allow a number of apparently hoax interrupts to occur before taking defensive action.

A hung device, while appearing to have work to do, might be failing to update its buffer descriptors. The driver should ascertain that it is not fooled by such repetitive requests.

On certain platforms, platform-specific bus drivers might be capable of identifying a persistently unclaimed interrupt and can disable the offending device. However, this relies on the driver's ability to identify the good interrupts and return the appropriate value. The driver should therefore return a DDI_INTR_UNCLAIMED result, unless it detects that the device legitimately asserted an interrupt (that is, the device actually requires the driver to do some useful work).

The legitimacy of other, more incidental, interrupts is much harder to certify. To this end, an interrupt expected flag is a useful tool for evaluating whether an interrupt is valid. Consider an interrupt such as descriptor free, which can be generated if, previously, all the device's descriptors had been allocated. If the driver detects that it has taken the last descriptor from the card, it can set an interrupt expected flag. If this flag is not set when the associated interrupt is delivered, it is suspicious.

Some informative interrupts might not be predictable, such as one indicating that a medium has become disconnected or frame sync has been lost. The easiest method of detecting whether such an interrupt is stuck is to mask this particular source on first occurrence until the next polling cycle.

If the interrupt occurs again while disabled, this should be considered a false interrupt. Some devices have interrupt status bits that can be read even if the mask register has disabled the associated source and might not be causing the interrupt. Driver designers can devise more appropriate algorithms specific to their devices.

Avoid looping on interrupt status bits indefinitely. Break such loops if none of the status bits set at the start of a pass requires any real work.

DMA Isolation

A defective PCI device might initiate an improper DMA transfer over the bus. This data transfer could corrupt good data previously delivered. A device that fails might generate a corrupt address that can contaminate memory not even belonging to this driver.

In systems with an IOMMU, a device can write only to pages mapped as writable for DMA. It is therefore important that DMA writes data only into pages that are owned by that driver instance and not shared with any other kernel structure. While the page in question is mapped as writable for DMA, the driver should be suspicious of data in that page, and the page must be unmapped from the IOMMU before it is passed beyond the driver, or before any validation of the data.

To guarantee that a whole aligned page is allocated, ddi_umem_alloc(9F) can be used, or multiple pages allocated and the memory below the first page boundary ignored. The size of an IOMMU page can be found using ddi_ptob(9F).

Alternatively, the driver can choose to copy the data into a safe part of memory before processing it. If this is done, the data must first be synchronized using ddi_dma_sync(9F).

Calls to ddi_dma_sync(9F) should specify SYNC_FOR_DEV before using DMA to transfer data to a device, and SYNC_FOR_CPU after using DMA to transfer data from the device to memory.

Some PCI devices are able to use dual address cycles (64-bit addresses) to avoid the IOMMU. This gives the device the potential to corrupt any region of main memory. Hardened device drivers must not attempt to use such a mode and should disable it.

Thread Interaction

Kernel panics in a device driver are often caused by unexpected interaction of kernel threads after a device failure. When a device fails, threads can interact in ways that the designer had not anticipated.

If processing routines terminate early, the condition variable waiters are blocked because an expected signal is never given. Attempting to inform other modules of the failure or handling unanticipated callbacks can result in undesirable thread interactions. Consider the sequence of mutex acquisition and relinquishing that can occur during device failures.

Threads that originate in an upstream STREAMS module can run into unfortunate paradoxes if used to return to that module unexpectedly. You might use alternative threads to handle exception messages. For instance, a wput procedure might use a read-side service routine to communicate an M_ERROR, rather than doing it directly with a read-side putnext.

A failing STREAMS device that cannot be quiesced during close (because of the fault) can generate an interrupt after the stream has been dismantled. The interrupt handler must not attempt to use a stale stream pointer to try to process the message.

Threats From Top-Down Requests

While protecting the system from defective hardware, the driver designer also needs to protect against driver misuse. Although the driver can assume that the kernel infrastructure is always correct (a trusted core), user requests passed to it can be potentially destructive.

For example, a user can request an action to be performed upon a user-supplied data block (M_IOCTL) that is smaller than that indicated in the control part of the message. The driver should never trust a user application.

The design should consider the construction of each type of ioctl that it can receive with a view to the potential harm that it could cause. The driver should make checks to be sure that it does not process malformed ioctls.

Adaptive Strategies

A driver can continue to provide a service with faulty hardware, attempting to work around the identified problem by using an alternative strategy for accessing the device. Given the unpredictability of broken hardware and the risk associated with additional design complexity, adaptive strategies are not always wise. At most, they should be limited to periodic interrupt polling and retry attempts. Periodically retrying the device lets the driver know when a device has recovered. Periodic polling can control the interrupt mechanism, after a driver has been forced to disable interrupts.

Ideally, a system always has an alternative device to provide a vital system service. Service multiplexors in kernel or user space offer the best method of maintaining system services when a device fails. Such practices are beyond the scope of this appendix.