Subversion Repositories HelenOS-doc

Rev

Rev 78 | Go to most recent revision | Details | Compare with Previous | Last modification | View Log | RSS feed

Rev Author Line No. Line
30 jermar 1
\chapter{Software}
2
\label{tools}
3
 
4
During the development of the HelenOS operating system, we came across
5
several types of software tools, programs, utilities and libraries.
6
Some of the tools were used to develop the system itself while other tools
7
were used to faciliate the development process. In some cases, we had a chance
8
to try out several versions of the same product. Sometimes the new versions
9
contained fixes for bugs we had discovered in previous versions thereof.
10
 
11
Another group of software we have used has been integrated into HelenOS
12
to fill gaps after functionality that the genuine HelenOS code did
13
not provide itself.
14
 
15
There is simply too much third party software that is somehow related to
16
HelenOS to be covered all. This chapter attempts to present our experience
78 jermar 17
with the key software tools, programs and libraries.
30 jermar 18
 
19
\section{Communication tools}
20
Although the developers know each other in person, the development, with the
21
exception of kernel camps, has been pretty much independent as far as locality
22
and time goes. In order to work effectively, we have established several communication
23
channels:
24
 
25
\begin{description}
26
\item [E-mail] --- We used this basic means of electronic communication for peer-to-peer
27
discussion in cases when the other person could not have been reached on-line at
28
the time his advice was needed or his attention was demanded. E-mail was also
29
used for contacting developers of third party software that we needed to talk to.
30
 
31
\item [Mailing list] --- As almost every open source project before us, also we opened
32
mailing list for technical discussion. The advantage of having a mailing list is
33
the fact that it enables multilateral discussions on several topics contemporarily,
34
without the need for all the participants be on-line or even at one place. We have kept
35
our first development mailing list closed to public so that it seemed natural to us
36
to use Czech as our communication language on the list since Czech, with one exception,
37
is our native language and all of us speak it very well. Besides all the advantages,
38
there are also disadvantages. First, communication over mailing list tends to be rather
39
slow, compared for instance to ICQ. Second, because of its implicit collective nature,
40
it sometimes tends to be so slow that an answer for a given question never comes.
41
 
42
Apart from the internal development mailing list, we have also used another mailing list
43
for commit log messages which proved handy in keeping developers informed about all changes in
44
the repository.
45
 
46
Finally, we have also established a public mailing list for communication
47
about general HelenOS topics in English.
48
 
49
\item [ICQ] --- Because we divided the whole project into smaller subprojects on which
50
only the maximum of two people out of six would work together, the need for communication
51
among all six people was significantly smaller than the need to communicate between the two
52
developers who tightly cooperated on a specific task. For this reason, we made the biggest
53
use of ICQ.
54
\end{description}
55
 
56
\section{Concurrent versions systems}
57
At the very beginning, when the SPARTAN kernel was being developed solely
58
by \JJ, there was not much sence in using any software for management of
59
concurrent versions. However, when the number of developers increased to six,
60
we immediately started to think of available solutions.
61
 
62
We have begun with CVS because it is probably the best known file concurrent
63
versions system. We have even had repository of HelenOS using CVS for a short time,
64
but when we learned about its weaknesses we sought another solution. There are two
65
weaknesses that have prevented us from using CVS:
66
 
67
\begin{itemize}
68
\item it is merely a file concurrent versions system (i.e. CVS is
69
good at managing versions of each separate file in the repository
70
but has no clue about the project's directory tree as a whole;
71
specifically renaming of a file while preserving its revision history
72
is next to impossible),
73
 
74
\item it lacks atomic commits (i.e. should your commit conflict with
75
another recent commit of another developer, CVS would not abort the whole operation
76
but render the repository inconsistent instead).
77
\end{itemize}
78
 
79
Being aware of these limitations, we decided to go with Subversion. Subversion
80
is, simply put, a redesigned CVS with all the limitations fixed. We were
81
already familiar with CVS so the switch to Subversion was pretty seamless.
82
 
83
As for Subversion itself, it has worked for us well and has met all our
84
expectations. Despite all its pros, there was a serious problem that
85
occurred sometime in the middle of the development process. Because of some locking
86
issues related to the default database backend (i.e. {\tt Berkeley DB}),
87
our Subversion repository put itself in a peculiar state in which it became
88
effectivelly inaccessible by any means of standard usage or administration.
89
To mitigate this problem, we had to manually delete orphaned file locks
90
and switch to backend called {\tt fsfs} which doesn't suffer this
91
problem.
92
 
93
Other than that, we are happy users of Subversion. The ability to switch
94
the entire working copy to particular revision is a great feature
95
for debugging. Once we tracked a bug three months into the past by
96
moving through revisions until we found the change that caused the bug.
97
 
98
\section{Web tools}
99
On our project website\cite{helenos}, we provided links to different
100
web utilities that either functioned to access our Subversion repository
101
or mailing list or provided another services:
102
 
103
\begin{description}
104
\item [Chora] is a part of the Horde framework and can be used to comfortably
105
browse Subversion repository from the web. We altered it a little bit to also
106
show number of commits per developer on our homepage.
107
 
33 jermar 108
\item [Whups] is another component of the Horde framework. It provides
30 jermar 109
feature request and bug tracking features. However, in the light of being rather
110
closed group of people, we used this tool only seldomly. On the other hand,
111
any possible beta tester of our operating system has had a chance to
112
submit bug reports.
113
 
114
\item [Mailman] is a web interface to the mailing list we utilized. It allows
115
to control subsriptions and search mailing list archives on-line.
116
\end{description}
117
 
118
\section{Third party components of HelenOS}
119
HelenOS itself contains third party software. In the first place, amd64 and ia32 architectures
78 jermar 120
make use of the GNU Grub boot loader. This software replaced the original limited boot loader
30 jermar 121
after the Kernel Camp 2005 when {\MD} had made HelenOS Multiboot specification compliant. Because of
122
Grub, HelenOS can be booted from several types of devices. More importantly, we use
123
Grub to load HelenOS userspace modules as well.
124
 
125
Another third-party piece of the HelenOS operating system is the userspace {\tt malloc()}.
126
Rather than porting our kernel slab allocator to userspace, we have chosen Doug Lea's public
127
domain {\tt dlmalloc} instead. This allocator could be easily integrated into our uspace tree
128
and has proven itself in other projects as well. Its derivative, {\tt ptmalloc}, has been part of the
129
GNU C library for some time. However, the version we are using is not optimized for SMP and multithreading.
130
We plan to eventually replace it with another allocator.
131
 
78 jermar 132
Next, the {\tt pci} userspace task is using the {\tt libpci} library. The
36 jermar 133
library was simplified and ported to HelenOS. Even though filesystem
134
calls were removed from the library, it still heavily depends on {\tt libc}.
135
By porting {\tt libpci} to HelenOS, we demonstrated that applications and libraries
136
are, given enough effort, portable to HelenOS.
137
 
78 jermar 138
Finally, we demonstrated the idea presented in the previous paragraph by porting
139
over 13 years old BSD game of {\tt tetris} to HelenOS. This particular version
140
of tetris looks almost the same both on other people's operating systems and on HelenOS.
141
Similar to {\tt libpci}, {\tt tetris} had to be modified in order to compile and run.
142
The filesystem calls were removed or replaced as well as references to terminal I/O
143
 calls.
144
 
30 jermar 145
\section{Build tools}
146
Assembler, linker and compiler are by all means the very focal point of attention
147
of all operating system projects. Quality of these tools influences
148
operating system performance and, what is more important, stability. HelenOS has
78 jermar 149
been tailored to build with GNU {\tt binutils}\cite{binutils} (i.e. the assembler and linker) and GNU~{\tt gcc}\cite{gcc}
30 jermar 150
(i.e. the compiler). There is only little chance that it could be compiled and
151
linked using some other tools unless those tools are compatible with the GNU build tools.
152
 
153
As our project declares support for five different processor architectures,
154
we needed to have five different flavors of the build utilities installed.
155
Interestingly, flavors of {\tt binutils} and {\tt gcc} for particular architecture
156
are not equal from the point of view of cross-binutils and cross-compiler installation.
157
All platforms except ia64 require only the {\tt binutils} package and the {\tt gcc} package
158
for the cross-tool to be built. On the other hand, ia64 requires also some excerpts from
159
the ia64-specific part of {\tt glibc}.
160
 
161
Formerly, the project could be compiled with almost any version of {\tt binutils} starting with 2.15
162
and {\tt gcc} starting with 2.95, but especially after we added partial thread local storage
163
support into our userspace layer, some architectures (e.g. mips32) will not compile even with {\tt gcc} 4.0.1
51 jermar 164
and demand {\tt gcc} 4.1.0 or newer.
30 jermar 165
 
166
As for the mips32 cross-compiler, {\OP} discovered a bug in {\tt gcc} (ticket \#23824) which caused {\tt gcc} to
167
incorrectly generate unaligned data access instructions (i.e. {\tt lwl}, {\tt lwr}, {\tt swl} and {\tt swr}).
168
 
169
As for the mips32 cross-binutils\footnote{It remains uninvestigated whether this problem also shows with other cross-tools.},
170
we observed that undefined symbols are not reported when we don't link using the standard target. We are still not
171
sure whether this was a bug --- {\tt binutils} developers just told us to use the standard target and then use
172
{\tt objcopy} to convert the ELF binary into requested output format.
173
 
174
\section{Virtual environments}
175
After the build tools, simulators, emulators and virtualizers were the second focal point
176
in our project. These invaluable programs really sped the code-compile-test cycle.
177
In some cases, they were, and still are, the only option to actually run HelenOS on certain
178
processor architectures, because real hardware was not available to us. Using virtual environment
33 jermar 179
for developing our system provided us with deterministic environment on which it is much easier to do
30 jermar 180
troubleshooting. Moreover, part of the simulators featured integrated debugging facilities.
181
Without them, a lot of bugs would remain unresolved or even go unnoticed.
182
 
42 jermar 183
Using several virtual environments for testing one architecture is well justified by the
184
fact that sometimes HelenOS would run on two and crash on third or vice versa. Sometimes
185
we found that it runs on real hardware but fails in a simulator. The opposite case was,
186
however, more common. Simply put, the more configurations, no matter whether real or virtual,
187
the better.
188
 
30 jermar 189
From one point of view, we have tested our system on eight different virtual environments:
190
 
191
\begin{itemize}
192
\item Bochs,
193
\item GXemul,
194
\item msim,
195
\item PearPC,
196
\item QEMU,
197
\item Simics,
198
\item Ski,
199
\item VMware.
200
\end{itemize}
201
 
202
From the second point of view, we have tested these programs by our operating system.
203
Because of the scope and uniqueness of this testing and because we did find some issues,
204
we want to dedicate some more space to what we have found.
205
 
206
\subsection{Bochs}
78 jermar 207
Bochs\cite{bochs} has been used to develop the SPARTAN kernel since its beginning in 2001.
30 jermar 208
It is capable of emulating ia32 machine and for some time also amd64.
209
Bochs is an emulator and thus the slowest from virtual environments capable
210
of simulating the same cathegory of hardware. On the other hand, it is extremely
211
portable, compared to much faster virtualizers and emulators using dynamic translation
212
of instructions. Lately, there have been some plans to develop or port dynamic translation
213
to Bochs brewing in its developer community.
214
 
215
The biggest virtue of Bochs is that it has traditionally supported SMP. For some time, Bochs
33 jermar 216
has been our only environment on which we could develop and test SMP code. Unfortunatelly,
30 jermar 217
the quality of SMP support in Bochs was different from version to version. Because of SMP
218
breakage in Bochs, we had to avoid some versions thereof. So far, Bochs versions 2.2.1 and 2.2.6
219
have been best in this regard.
220
 
221
Our project has not only used Bochs. We also helped to identify some SMP related problems
222
and {\OP} from our team has discovered and also fixed a bug in FXSAVE and FXRSTOR emulation
223
(patch \#1282033).
224
 
225
Bochs has some debugging facilities but those have been very impractical and broken
42 jermar 226
in SMP mode. Moreover, it is possible to use the GNU debugger {\tt gbd} to connect to running
227
simulation, but this has also proven not very useful as we often needed to debug
228
problems that existed only in multiprocessor configurations, which {\tt gdb}
229
does not understand.
30 jermar 230
 
231
\subsection{GXemul}
78 jermar 232
GXemul\cite{gxemul} is an emulator of several processor architectures. Nevertheless, we have
30 jermar 233
used it only for mips32 emulation in both little-endian and big-endian modes.
234
It seems to be pretty featurefull and evolving but we don't use all its functionality.
235
GXemul is very user friendly and has debugging features. It is more realistic
236
than msim. However, our newly introduced TLS support triggered a bug in the {\tt rdhwr}
237
instruction emulation while msim functioned as expected. Fortunatelly, the author
238
of GXemul is very cooperative and has fixed the problem for future versions as well as
239
provided a quick hack for the old version.
240
 
241
\subsection{msim}
78 jermar 242
msim\cite{msim} has been our first mips32 simulator. It simulates 32-bit side of R4000 processor.
30 jermar 243
Its simulated environment is not very realistic, but the processor simulation
244
is good enough for operating system development. In this regard, the simulator is
245
comparable to HP's ia64 simulator Ski. Another similar aspect of these two is
246
relatively strong debugger.
247
 
248
Msim has been developed on the same alma mater as our own project.
249
All members of our team know this program from operating system courses.
250
Curiously, this simulator contained the biggest number of defects and inaccuracies
251
that we have ever discovered in a simulator.  Fortunately, all of them have been
252
eventually fixed.
253
 
254
\subsection{PearPC}
78 jermar 255
PearPC\cite{pearpc} is the only emulator on which we have run ppc32 port of HelenOS. It has
30 jermar 256
no debugging features, but fortunatelly its sources are available under
257
an open source license. This enabled {\OP} and {\MD} to alter its sources
258
in a way that this modified version allowed some basic debugging.
259
 
260
\subsection{QEMU}
78 jermar 261
QEMU\cite{qemu} emulates several processor architectures. We have used it to emulate
30 jermar 262
ia32 and amd64. It can simulate SMP, but contrary to Bochs, it uses dynamic
263
translation of emulated instructions and performs much better because of
264
that.
265
 
32 jermar 266
This emulator seemed to realistically emulate the {\tt hlt} instruction,
267
which was nice for those of us who use notebooks as their development
268
machine.
269
 
42 jermar 270
Similar to Bochs, QEMU simulation can be aided by {\tt gdb}. Debugging
271
with {\tt gdb} can be pretty comfortable\footnote{Especially when the kernel is
272
compiled with {\tt -g3}.} until one needs to debug a SMP kernel running on multiple
273
processors.
274
 
30 jermar 275
\subsection{Simics}
78 jermar 276
Virtutech's Simics\cite{simics} simulator can be compared to a Swiss-army knife for operating system debugging.
32 jermar 277
This proprietary piece of software was available to us under an academic license for free.
278
 
279
Simics can be set to simulate many different configurations of many different machines.
280
It has the most advanced debugging features we have ever seen. To highlight some, its
281
memory access tracing ability has been really helpfull to us. During device driver
282
development, we appreciated the possibility to turn logging of the devices to a specified
283
verbosity.
284
 
148 palkovsky 285
We used it to test and develop amd64 and ia32 architectures in SMP mode and mips32 architecture in UP mode. Simics emulates the 4Kc processor on the MIPS architecture.
286
Unfortunately, this processor does not have an exception Reserved Instruction, which
287
makes it unusable in an environment with programs using thread local storage.
32 jermar 288
 
289
Regardless of its invaluable qualities, it has still contained bugs. One of the most
290
serious was bug with ticket \#3351. {\OP} discovered that its BIOS rewrites kernel memory
291
during application processors start. Another bugs found were related to amd64 and mips32.
292
As for amd64, Simics did not report general protection fault when {\tt EFER.NXE} was 0 and a non-executable
293
page was found (\#4214). As for mips32, Simics misemulated {\tt MSUB} and {\tt MSUBU} instructions.
294
 
30 jermar 295
\subsection{Ski}
78 jermar 296
The ia64 port of HelenOS has been developed and debugged on the HP's IA-64 Ski\cite{ski} simulator.
33 jermar 297
Ski is just an Itanium processor simulator and as such does not simulate a real machine. In fact, there
298
is no firmware and no configuration tables (e.g. memory map) present in Ski! On the other hand, the missing parts can be supplied externally\footnote{This
299
is actually how Linux runs in this simulator.}. The simulator provides means of interaction with
300
host system devices via Simulator SystemCalls (SSC). The simulator itself has graphical interface
301
with pretty powerful, but not as good as those of Simics, debugging facilities.
302
 
303
Ski is a proprietary program with no source code available. Its binaries are available
304
for free under a non-free license. It comes packaged with insufficient documentation
305
which makes the development pretty problematic. For instance, there is no public documentation
306
of all the SSC's. All one can do is to look into Linux/ia64-Ski port, which was written by the
307
same people as Ski, and use it as a refernce. We had to look into Linux once more when our kernel
308
started to fail in some memory-intensive stress tests. In fact, the problem was that the tests
309
hit the IA-32 legacy videoram area. We fixed the problem, in the light of absence of any memory map, by blacklisting
310
this piece of memory to our frame allocator.
311
 
312
The way HelenOS is booted on Ski is by simply loading its ELF image
313
and jumping to it. The ELF header contains two fields describing where and how to load the program image into memory:
314
VMA and LMA. VMA\footnote{Virtual Memory Address} is an address where the program's segment gets mapped in virtual memory.
315
LMA\footnote{Load Memory Address} is the physical address where the segment is loaded in memory. {\JV} discovered
316
that Ski confuses VMA and LMA. This, what we believe to be a bug in Ski, has not shown in Linux since Linux always has
317
LMA equal to VMA. People from the Ski mailing list had tried to help us but our repeated problem report didn't
318
make it far enough for the HP to fix or at least clarify the issue. Finally, we adopted a workaround implemented by {\JJ}
319
that simply swaps LMA and the program entry point in the kernel ELF image.
320
 
78 jermar 321
\subsection{VMware} VMware\cite{vmware} is the only virtualizer we have used in
42 jermar 322
HelenOS development. It virtualizes the ia32 host machine. Since VMware
323
version 5.5, we made use of its possibility to run the guest system
324
(i.e. HelenOS) on multiple processors. VMware has no support for
325
debugging but is very useful for compatibility and regression testing
326
because it's closest to the real hardware. VMware, being a virtualizer,
327
is also the fastest of all the virtual environments we have utilized.
30 jermar 328
 
33 jermar 329