1 (How to avoid) Botching up ioctls
2 =================================
4 From: http://blog.ffwll.ch/2013/11/botching-up-ioctls.html
6 By: Daniel Vetter, Copyright © 2013 Intel Corporation
8 One clear insight kernel graphics hackers gained in the past few years is that
9 trying to come up with a unified interface to manage the execution units and
10 memory on completely different GPUs is a futile effort. So nowadays every
11 driver has its own set of ioctls to allocate memory and submit work to the GPU.
12 Which is nice, since there's no more insanity in the form of fake-generic, but
13 actually only used once interfaces. But the clear downside is that there's much
14 more potential to screw things up.
16 To avoid repeating all the same mistakes again I've written up some of the
17 lessons learned while botching the job for the drm/i915 driver. Most of these
18 only cover technicalities and not the big-picture issues like what the command
19 submission ioctl exactly should look like. Learning these lessons is probably
20 something every GPU driver has to do on its own.
26 First the prerequisites. Without these you have already failed, because you
27 will need to add a 32-bit compat layer:
29 * Only use fixed sized integers. To avoid conflicts with typedefs in userspace
30 the kernel has special types like __u32, __s64. Use them.
32 * Align everything to the natural size and use explicit padding. 32-bit
33 platforms don't necessarily align 64-bit values to 64-bit boundaries, but
34 64-bit platforms do. So we always need padding to the natural size to get
37 * Pad the entire struct to a multiple of 64-bits if the structure contains
38 64-bit types - the structure size will otherwise differ on 32-bit versus
39 64-bit. Having a different structure size hurts when passing arrays of
40 structures to the kernel, or if the kernel checks the structure size, which
41 e.g. the drm core does.
43 * Pointers are __u64, cast from/to a uintprt_t on the userspace side and
44 from/to a void __user * in the kernel. Try really hard not to delay this
45 conversion or worse, fiddle the raw __u64 through your code since that
46 diminishes the checking tools like sparse can provide. The macro
47 u64_to_user_ptr can be used in the kernel to avoid warnings about integers
48 and pointres of different sizes.
54 With the joys of writing a compat layer avoided we can take a look at the basic
55 fumbles. Neglecting these will make backward and forward compatibility a real
56 pain. And since getting things wrong on the first attempt is guaranteed you
57 will have a second iteration or at least an extension for any given interface.
59 * Have a clear way for userspace to figure out whether your new ioctl or ioctl
60 extension is supported on a given kernel. If you can't rely on old kernels
61 rejecting the new flags/modes or ioctls (since doing that was botched in the
62 past) then you need a driver feature flag or revision number somewhere.
64 * Have a plan for extending ioctls with new flags or new fields at the end of
65 the structure. The drm core checks the passed-in size for each ioctl call
66 and zero-extends any mismatches between kernel and userspace. That helps,
67 but isn't a complete solution since newer userspace on older kernels won't
68 notice that the newly added fields at the end get ignored. So this still
69 needs a new driver feature flags.
71 * Check all unused fields and flags and all the padding for whether it's 0,
72 and reject the ioctl if that's not the case. Otherwise your nice plan for
73 future extensions is going right down the gutters since someone will submit
74 an ioctl struct with random stack garbage in the yet unused parts. Which
75 then bakes in the ABI that those fields can never be used for anything else
76 but garbage. This is also the reason why you must explicitly pad all
77 structures, even if you never use them in an array - the padding the compiler
78 might insert could contain garbage.
80 * Have simple testcases for all of the above.
86 Nowadays we don't have any excuse left any more for drm drivers being neat
87 little root exploits. This means we both need full input validation and solid
88 error handling paths - GPUs will die eventually in the oddmost corner cases
91 * The ioctl must check for array overflows. Also it needs to check for
92 over/underflows and clamping issues of integer values in general. The usual
93 example is sprite positioning values fed directly into the hardware with the
94 hardware just having 12 bits or so. Works nicely until some odd display
95 server doesn't bother with clamping itself and the cursor wraps around the
98 * Have simple testcases for every input validation failure case in your ioctl.
99 Check that the error code matches your expectations. And finally make sure
100 that you only test for one single error path in each subtest by submitting
101 otherwise perfectly valid data. Without this an earlier check might reject
102 the ioctl already and shadow the codepath you actually want to test, hiding
103 bugs and regressions.
105 * Make all your ioctls restartable. First X really loves signals and second
106 this will allow you to test 90% of all error handling paths by just
107 interrupting your main test suite constantly with signals. Thanks to X's
108 love for signal you'll get an excellent base coverage of all your error
109 paths pretty much for free for graphics drivers. Also, be consistent with
110 how you handle ioctl restarting - e.g. drm has a tiny drmIoctl helper in its
111 userspace library. The i915 driver botched this with the set_tiling ioctl,
112 now we're stuck forever with some arcane semantics in both the kernel and
115 * If you can't make a given codepath restartable make a stuck task at least
116 killable. GPUs just die and your users won't like you more if you hang their
117 entire box (by means of an unkillable X process). If the state recovery is
118 still too tricky have a timeout or hangcheck safety net as a last-ditch
119 effort in case the hardware has gone bananas.
121 * Have testcases for the really tricky corner cases in your error recovery code
122 - it's way too easy to create a deadlock between your hangcheck code and
126 Time, Waiting and Missing it
127 ----------------------------
129 GPUs do most everything asynchronously, so we have a need to time operations and
130 wait for outstanding ones. This is really tricky business; at the moment none of
131 the ioctls supported by the drm/i915 get this fully right, which means there's
132 still tons more lessons to learn here.
134 * Use CLOCK_MONOTONIC as your reference time, always. It's what alsa, drm and
135 v4l use by default nowadays. But let userspace know which timestamps are
136 derived from different clock domains like your main system clock (provided
137 by the kernel) or some independent hardware counter somewhere else. Clocks
138 will mismatch if you look close enough, but if performance measuring tools
139 have this information they can at least compensate. If your userspace can
140 get at the raw values of some clocks (e.g. through in-command-stream
141 performance counter sampling instructions) consider exposing those also.
143 * Use __s64 seconds plus __u64 nanoseconds to specify time. It's not the most
144 convenient time specification, but it's mostly the standard.
146 * Check that input time values are normalized and reject them if not. Note
147 that the kernel native struct ktime has a signed integer for both seconds
148 and nanoseconds, so beware here.
150 * For timeouts, use absolute times. If you're a good fellow and made your
151 ioctl restartable relative timeouts tend to be too coarse and can
152 indefinitely extend your wait time due to rounding on each restart.
153 Especially if your reference clock is something really slow like the display
154 frame counter. With a spec lawyer hat on this isn't a bug since timeouts can
155 always be extended - but users will surely hate you if their neat animations
156 starts to stutter due to this.
158 * Consider ditching any synchronous wait ioctls with timeouts and just deliver
159 an asynchronous event on a pollable file descriptor. It fits much better
160 into event driven applications' main loop.
162 * Have testcases for corner-cases, especially whether the return values for
163 already-completed events, successful waits and timed-out waits are all sane
164 and suiting to your needs.
167 Leaking Resources, Not
168 ----------------------
170 A full-blown drm driver essentially implements a little OS, but specialized to
171 the given GPU platforms. This means a driver needs to expose tons of handles
172 for different objects and other resources to userspace. Doing that right
173 entails its own little set of pitfalls:
175 * Always attach the lifetime of your dynamically created resources to the
176 lifetime of a file descriptor. Consider using a 1:1 mapping if your resource
177 needs to be shared across processes - fd-passing over unix domain sockets
178 also simplifies lifetime management for userspace.
180 * Always have O_CLOEXEC support.
182 * Ensure that you have sufficient insulation between different clients. By
183 default pick a private per-fd namespace which forces any sharing to be done
184 explicitly. Only go with a more global per-device namespace if the objects
185 are truly device-unique. One counterexample in the drm modeset interfaces is
186 that the per-device modeset objects like connectors share a namespace with
187 framebuffer objects, which mostly are not shared at all. A separate
188 namespace, private by default, for framebuffers would have been more
191 * Think about uniqueness requirements for userspace handles. E.g. for most drm
192 drivers it's a userspace bug to submit the same object twice in the same
193 command submission ioctl. But then if objects are shareable userspace needs
194 to know whether it has seen an imported object from a different process
195 already or not. I haven't tried this myself yet due to lack of a new class
196 of objects, but consider using inode numbers on your shared file descriptors
197 as unique identifiers - it's how real files are told apart, too.
198 Unfortunately this requires a full-blown virtual filesystem in the kernel.
204 Not every problem needs a new ioctl:
206 * Think hard whether you really want a driver-private interface. Of course
207 it's much quicker to push a driver-private interface than engaging in
208 lengthy discussions for a more generic solution. And occasionally doing a
209 private interface to spearhead a new concept is what's required. But in the
210 end, once the generic interface comes around you'll end up maintainer two
211 interfaces. Indefinitely.
213 * Consider other interfaces than ioctls. A sysfs attribute is much better for
214 per-device settings, or for child objects with fairly static lifetimes (like
215 output connectors in drm with all the detection override attributes). Or
216 maybe only your testsuite needs this interface, and then debugfs with its
217 disclaimer of not having a stable ABI would be better.
219 Finally, the name of the game is to get it right on the first attempt, since if
220 your driver proves popular and your hardware platforms long-lived then you'll
221 be stuck with a given ioctl essentially forever. You can try to deprecate
222 horrible ioctls on newer iterations of your hardware, but generally it takes
223 years to accomplish this. And then again years until the last user able to
224 complain about regressions disappears, too.