| 1 |
neysx 05/07/28 13:21:06 |
| 2 |
|
| 3 |
Added: xml/htdocs/doc/en/articles hardware-stability-p1.xml |
| 4 |
hardware-stability-p2.xml |
| 5 |
Log: |
| 6 |
#100524 xmlified hardware stability articles |
| 7 |
|
| 8 |
Revision Changes Path |
| 9 |
1.1 xml/htdocs/doc/en/articles/hardware-stability-p1.xml |
| 10 |
|
| 11 |
file : http://www.gentoo.org/cgi-bin/viewcvs.cgi/xml/htdocs/doc/en/articles/hardware-stability-p1.xml?rev=1.1&content-type=text/x-cvsweb-markup&cvsroot=gentoo |
| 12 |
plain: http://www.gentoo.org/cgi-bin/viewcvs.cgi/xml/htdocs/doc/en/articles/hardware-stability-p1.xml?rev=1.1&content-type=text/plain&cvsroot=gentoo |
| 13 |
|
| 14 |
Index: hardware-stability-p1.xml |
| 15 |
=================================================================== |
| 16 |
<?xml version="1.0" encoding="UTF-8"?> |
| 17 |
<!DOCTYPE guide SYSTEM "/dtd/guide.dtd"> |
| 18 |
<!-- $Header: /var/cvsroot/gentoo/xml/htdocs/doc/en/articles/hardware-stability-p1.xml,v 1.1 2005/07/28 13:21:06 neysx Exp $ --> |
| 19 |
|
| 20 |
<guide link="/doc/en/articles/hardware-stability-p1.xml" lang="en"> |
| 21 |
<title>Linux hardware stability guide, Part 1</title> |
| 22 |
|
| 23 |
<author title="Author"> |
| 24 |
<mail link="drobbins@g.o">Daniel Robbins</mail> |
| 25 |
</author> |
| 26 |
<author title="Editor"> |
| 27 |
<mail link="jackdark@×××××.com">Joshua Saddler</mail> |
| 28 |
</author> |
| 29 |
|
| 30 |
<abstract> |
| 31 |
In this article, Daniel Robbins shows you how to diagnose and fix CPU |
| 32 |
flakiness, as well as how to test your RAM for defects. By the end of this |
| 33 |
article, you'll have the skills to ensure that your Linux system is as stable |
| 34 |
as it possibly can be. |
| 35 |
</abstract> |
| 36 |
|
| 37 |
<!-- The original version of this article was first published on IBM |
| 38 |
developerWorks, and is property of Westtech Information Services. This |
| 39 |
document is an updated version of the original article, and contains |
| 40 |
various improvements made by the Gentoo Linux Documentation team --> |
| 41 |
|
| 42 |
<version>1.0</version> |
| 43 |
<date>2005-07-28</date> |
| 44 |
|
| 45 |
<chapter> |
| 46 |
<title>CPU troubleshooting</title> |
| 47 |
<section> |
| 48 |
<body> |
| 49 |
|
| 50 |
<note> |
| 51 |
The original version of this article was first published on IBM developerWorks, |
| 52 |
and is property of Westtech Information Services. This document is an updated |
| 53 |
version of the original article, and contains various improvements made by the |
| 54 |
Gentoo Linux Documentation team. |
| 55 |
</note> |
| 56 |
|
| 57 |
<p> |
| 58 |
Many of us in the Linux world have been bitten by nasty hardware problems. How |
| 59 |
many of us have set up a Linux box, installed our favorite distribution, |
| 60 |
compiled and installed some additional apps, and gotten everything working |
| 61 |
perfectly only to find that our new system has an (argh!) fatal hardware bug? |
| 62 |
Whether the symptoms are random segmentation faults, data corruption, hard |
| 63 |
locks, or lost data is irrelevant -- the hardware glitch effectively makes our |
| 64 |
normally reliable Linux operating system barely able to stay afloat. In this |
| 65 |
article, we'll take an in-depth look at how to detect flaky CPUs and RAM -- |
| 66 |
allowing you to replace the defective parts before they do some serious |
| 67 |
damage. |
| 68 |
</p> |
| 69 |
|
| 70 |
<p> |
| 71 |
If you're experiencing instability problems and suspect they are hardware |
| 72 |
related, I encourage you to test both your CPU and memory to ensure that |
| 73 |
they're working OK. However, even if you haven't experienced these problems, |
| 74 |
it's still a good idea to perform these CPU and memory tests. In doing so, you |
| 75 |
may detect a hardware problem that could have bitten you at an inopportune |
| 76 |
time, something that could have caused data loss or hours of frustration in a |
| 77 |
frantic search for the source of the problem. The proper, proactive application |
| 78 |
of these techniques can help you to avoid a lot of headaches, and if your |
| 79 |
system passes the tests, you'll have the peace of mind that your system is up |
| 80 |
to spec. |
| 81 |
</p> |
| 82 |
|
| 83 |
</body> |
| 84 |
</section> |
| 85 |
<section> |
| 86 |
<title>CPU issues</title> |
| 87 |
<body> |
| 88 |
|
| 89 |
<p> |
| 90 |
If you have a horribly defective CPU, your machine may be unable to boot Linux |
| 91 |
or may only run for a few minutes before locking up. CPUs in this ragged state |
| 92 |
are easy to diagnose as defective because the symptoms are so obvious. But |
| 93 |
there are more subtle CPU defects that aren't so easy to detect; generally, the |
| 94 |
less obvious errors are the ones that cause machines to either lock up every |
| 95 |
now and then for no apparent reason, or cause certain processes to die |
| 96 |
unexpectedly. Most CPU instabilities can be triggered by "exercising" the CPU |
| 97 |
-- giving it a bunch of work to do, causing it to heat up and possibly flake |
| 98 |
out. Let's look at some ways to stress-test the CPU. |
| 99 |
</p> |
| 100 |
|
| 101 |
<p> |
| 102 |
You may be surprised to hear that one of the best tests of CPU stability is |
| 103 |
built in to Linux -- the kernel compile. The gcc compiler is a great tool for |
| 104 |
testing general CPU stability, and a kernel build uses gcc a whole lot. By |
| 105 |
creating and running the following script from your <path>/usr/src/linux</path> |
| 106 |
directory, you can give your machine an industrial-strength kernel compile |
| 107 |
stress test: |
| 108 |
</p> |
| 109 |
|
| 110 |
<pre caption="The cpubuild script"> |
| 111 |
#!/bin/bash |
| 112 |
make dep |
| 113 |
while [ "foo" = "foo" ] |
| 114 |
do |
| 115 |
make clean |
| 116 |
make -j2 bzImage |
| 117 |
if [ $? -ne 0 ] |
| 118 |
then |
| 119 |
echo OUCH OUCH OUCH OUCH |
| 120 |
exit 1 |
| 121 |
fi |
| 122 |
done |
| 123 |
</pre> |
| 124 |
|
| 125 |
<p> |
| 126 |
You'll notice that this script <e>repeatedly</e> compiles the kernel. The |
| 127 |
reason for this is simple -- some CPUs have intermittent glitches, allowing |
| 128 |
them to compile the kernel perfectly 95% of the time, but causing the kernel |
| 129 |
compile to bomb out every now and then. Normally, this is because it may take |
| 130 |
five or more kernel compiles before the processor heats up to the point where |
| 131 |
it becomes unstable. |
| 132 |
</p> |
| 133 |
|
| 134 |
<p> |
| 135 |
In the above script, make sure to adjust the <c>-j</c> option so that the |
| 136 |
number following it is one greater than the number of CPUs in your system; in |
| 137 |
other words, use "2" for uniprocessors, "3" for dual-processors, etc. The |
| 138 |
<c>-j</c> option tells <c>make</c> to build the kernel in parallel, ensuring |
| 139 |
that there's always at least one gcc process on deck after each source file is |
| 140 |
compiled -- ensuring that the stress on your CPU is maximized. If your Linux |
| 141 |
box is going to be unused for the afternoon, go ahead and run this script, and |
| 142 |
let the machine recompile the kernel for a few hours. |
| 143 |
</p> |
| 144 |
|
| 145 |
</body> |
| 146 |
</section> |
| 147 |
<section> |
| 148 |
<title>Possible CPU problems</title> |
| 149 |
<body> |
| 150 |
|
| 151 |
<p> |
| 152 |
If the script runs perfectly for several hours, congratulations! Your CPU has |
| 153 |
passed the first test. However, it's possible that the above script dies |
| 154 |
unexpectedly. How do you know you're having a CPU problem as opposed to |
| 155 |
something else? Well, if gcc spat out an error like this, then there's a very |
| 156 |
good possibility that your CPU is defective: |
| 157 |
</p> |
| 158 |
|
| 159 |
<pre caption="GCC error"> |
| 160 |
gcc: Internal compiler error: program cc1 got fatal signal 11 |
| 161 |
</pre> |
| 162 |
|
| 163 |
<p> |
| 164 |
At this point, you have about three possibilities as to the state of your CPU: |
| 165 |
</p> |
| 166 |
|
| 167 |
<ul> |
| 168 |
<li> |
| 169 |
If you type <c>make bzImage</c> to resume the kernel compilation, and the |
| 170 |
compiler dies on the exact same file, keep typing <c>make bzImage</c> over |
| 171 |
and over again. If after about ten tries the build process continues to die |
| 172 |
on this particular file, then the problem is most likely caused by a (rare) |
| 173 |
gcc compiler bug that's being triggered by this particular source file, |
| 174 |
rather than a flaky CPU. However, these days, gcc is quite stable, so this |
| 175 |
isn't likely to happen. |
| 176 |
</li> |
| 177 |
<li> |
| 178 |
If you type <c>make bzImage</c> to resume kernel compilation, and you get |
| 179 |
another signal 11 a little bit later, then your CPU is most likely on its |
| 180 |
last legs. |
| 181 |
</li> |
| 182 |
<li> |
| 183 |
If you type <c>make bzImage</c> to resume kernel compilation and the kernel |
| 184 |
compiles successfully, this doesn't mean that your CPU is OK. Normally, |
| 185 |
this means that your CPU glitch only shows up every now and then, normally |
| 186 |
only when the CPU rises above a certain temperature (a CPU will get hotter |
| 187 |
when it is being used for an extended period of time, and may take several |
| 188 |
kernel compiles to get to that critical point). |
| 189 |
</li> |
| 190 |
</ul> |
| 191 |
|
| 192 |
</body> |
| 193 |
</section> |
| 194 |
<section> |
| 195 |
<title>Rescuing your CPU</title> |
| 196 |
<body> |
| 197 |
|
| 198 |
<p> |
| 199 |
If your CPU is experiencing random intermittent errors when placed under heavy |
| 200 |
load, it's possible that your CPU isn't defective at all -- maybe it simply |
| 201 |
isn't being cooled properly. Here are some things that you can check: |
| 202 |
</p> |
| 203 |
|
| 204 |
<ul> |
| 205 |
<li>Is your CPU fan plugged in?</li> |
| 206 |
<li>Is it relatively dust-free?</li> |
| 207 |
<li> |
| 208 |
Does the fan actually spin (and spin at the proper speed) when the power is |
| 209 |
on? |
| 210 |
</li> |
| 211 |
<li>Is the heat sink seated properly on the CPU?</li> |
| 212 |
<li>Is there thermal grease between the CPU and the heat sink?</li> |
| 213 |
<li>Does your case have adequate ventilation?</li> |
| 214 |
</ul> |
| 215 |
|
| 216 |
|
| 217 |
|
| 218 |
1.1 xml/htdocs/doc/en/articles/hardware-stability-p2.xml |
| 219 |
|
| 220 |
file : http://www.gentoo.org/cgi-bin/viewcvs.cgi/xml/htdocs/doc/en/articles/hardware-stability-p2.xml?rev=1.1&content-type=text/x-cvsweb-markup&cvsroot=gentoo |
| 221 |
plain: http://www.gentoo.org/cgi-bin/viewcvs.cgi/xml/htdocs/doc/en/articles/hardware-stability-p2.xml?rev=1.1&content-type=text/plain&cvsroot=gentoo |
| 222 |
|
| 223 |
Index: hardware-stability-p2.xml |
| 224 |
=================================================================== |
| 225 |
<?xml version="1.0" encoding="UTF-8"?> |
| 226 |
<!DOCTYPE guide SYSTEM "/dtd/guide.dtd"> |
| 227 |
<!-- $Header: /var/cvsroot/gentoo/xml/htdocs/doc/en/articles/hardware-stability-p2.xml,v 1.1 2005/07/28 13:21:06 neysx Exp $ --> |
| 228 |
|
| 229 |
<guide link="/doc/en/articles/hardware-stability-p2.xml" lang="en"> |
| 230 |
<title>Linux hardware stability guide, Part 2</title> |
| 231 |
|
| 232 |
<author title="Author"> |
| 233 |
<mail link="drobbins@g.o">Daniel Robbins</mail> |
| 234 |
</author> |
| 235 |
<author title="Editor"> |
| 236 |
<mail link="jackdark@×××××.com">Joshua Saddler</mail> |
| 237 |
</author> |
| 238 |
|
| 239 |
<abstract> |
| 240 |
In this article, Daniel Robbins shares his experiences in getting his NVIDIA |
| 241 |
TNT graphics card working under Linux using NVIDIA's accelerated drivers. As he |
| 242 |
does, he'll show you how to diagnose and fix IRQ and PCI latency timer issues |
| 243 |
-- techniques you can use to ensure that your systems don't experience |
| 244 |
lock-ups, inconsistent behavior, or data loss. |
| 245 |
</abstract> |
| 246 |
|
| 247 |
<!-- The original version of this article was first published on IBM |
| 248 |
developerWorks, and is property of Westtech Information Services. This |
| 249 |
document is an updated version of the original article, and contains |
| 250 |
various improvements made by the Gentoo Linux Documentation team --> |
| 251 |
|
| 252 |
<version>1.0</version> |
| 253 |
<date>2005-07-28</date> |
| 254 |
|
| 255 |
<chapter> |
| 256 |
<title>Drivers</title> |
| 257 |
<section> |
| 258 |
<title>The many causes of instability</title> |
| 259 |
<body> |
| 260 |
|
| 261 |
<note> |
| 262 |
The original version of this article was first published on IBM developerWorks, |
| 263 |
and is property of Westtech Information Services. This document is an updated |
| 264 |
version of the original article, and contains various improvements made by the |
| 265 |
Gentoo Linux Documentation team. |
| 266 |
</note> |
| 267 |
|
| 268 |
<p> |
| 269 |
A stability problem is often not caused by defective hardware, but by improper |
| 270 |
hardware configuration or flaky drivers. My experience in this area began when |
| 271 |
I tried to get Linux working on my Diamond Viper V550, an NVIDIA TNT-based AGP |
| 272 |
card, using NVIDIA's own accelerated drivers. |
| 273 |
</p> |
| 274 |
|
| 275 |
<p> |
| 276 |
NVIDIA has their own display drivers for Linux, a collaboration between NVIDIA, |
| 277 |
SGI, and VA Linux. These drivers have a lot of advantages over the standard |
| 278 |
2d-only NVIDIA drivers included with Xfree86 4.0. For one, they have full |
| 279 |
accelerated 3D support. Even better, they feature an official OpenGL 1.2 |
| 280 |
implementation, rather than just an enhanced version of Mesa. So, all in all, |
| 281 |
these accelerated drivers are the ones you want to be using if you own an |
| 282 |
NVIDIA-based graphics card, at least in theory. My attempt to get them working |
| 283 |
properly turned out to be an excellent learning experience, to say the least. |
| 284 |
</p> |
| 285 |
|
| 286 |
<p> |
| 287 |
After I installed the accelerated Linux NVIDIA drivers (see <uri |
| 288 |
link="#resources">Resources</uri> later in this article), I started up Xfree86 |
| 289 |
and began playing around with all my 3D applications, now wonderfully |
| 290 |
accelerated as they should be. Up until that point, I had had to reboot into |
| 291 |
Windows NT in order to take advantage of 3D acceleration. Now, while I don't |
| 292 |
mind NT, having to reboot to use 3D apps was somewhat annoying, and I was glad |
| 293 |
to have one less reason to leave Linux and reboot my machine. However, after |
| 294 |
playing around for an hour or so, I experienced a fatal setback to my Linux 3D |
| 295 |
aspirations -- my machine locked up. My mouse simply stopped moving and the |
| 296 |
screen froze, and I had to reboot my system. |
| 297 |
</p> |
| 298 |
|
| 299 |
<p> |
| 300 |
Yes, I was having some kind of stability problem. But I didn't know exactly |
| 301 |
what was causing the problem. Did I have flaky hardware, or was the card |
| 302 |
misconfigured? Or maybe it was a problem with the driver -- did it not like my |
| 303 |
VIA KT133-based Athlon motherboard? Whatever the problem, I wanted to resolve |
| 304 |
it quickly. In this article, I'm going to share with you the procedure that I |
| 305 |
went through to fix my hardware stability problem. Although you may not be |
| 306 |
struggling with exactly the same issue, the steps that I used to diagnose and |
| 307 |
(mostly) fix the problem are general in nature and applicable to many different |
| 308 |
types of Linux hardware problems. |
| 309 |
</p> |
| 310 |
|
| 311 |
</body> |
| 312 |
</section> |
| 313 |
<section> |
| 314 |
<title>First, the hardware</title> |
| 315 |
<body> |
| 316 |
|
| 317 |
<p> |
| 318 |
The first thing that crossed my mind was that I might have flaky or |
| 319 |
under-cooled hardware. On the one hand, my Diamond Viper V550 seemed to have no |
| 320 |
problems under Windows NT. On the other hand, maybe Linux was somehow pushing |
| 321 |
the chip harder and triggering heat-related lock-ups. My V550 did get |
| 322 |
<e>extremely</e> hot, and its OEM heatsink seemed at best barely adequate. The |
| 323 |
combination of the lock-ups and the fact that this card was being marginally |
| 324 |
cooled convinced me to head over to PC Power and Cooling (see <uri |
| 325 |
link="#resources">Resources</uri>) to purchase a mini integrated heatsink/fan |
| 326 |
for my V550. |
| 327 |
</p> |
| 328 |
|
| 329 |
<p> |
| 330 |
So, I received my Video Cool, popped off the OEM heatsink on the video card |
| 331 |
(voiding the warranty), cleaned off the TNT chip and affixed the Video Cool to |
| 332 |
the top of the chip. Verdict? My video card didn't get extremely hot anymore, |
| 333 |
but the lockups continued. The lesson I learned from this particular experience |
| 334 |
is this -- if you ensure that your system is adequately cooled to begin with, |
| 335 |
you'll never need to worry about components malfunctioning due to inadequate |
| 336 |
cooling. This in itself is a good reason to invest some time and effort in |
| 337 |
making sure that your workstations and servers run coolly. Now that I had taken |
| 338 |
care of the heat issue, I knew that the lock-ups were most likely not due to |
| 339 |
flaky hardware, and I began to look elsewhere. |
| 340 |
</p> |
| 341 |
|
| 342 |
</body> |
| 343 |
</section> |
| 344 |
<section> |
| 345 |
<title>New drivers -- and a possible solution?</title> |
| 346 |
<body> |
| 347 |
|
| 348 |
<p> |
| 349 |
I partly suspected that NVIDIA's drivers were themselves the cause of the |
| 350 |
problem. Fortunately , a new version of the drivers had just been released, so |
| 351 |
I immediately upgraded in the hope that this would solve my stability problem. |
| 352 |
Unfortunately, it didn't, and after checking with others on the #nvidia channel |
| 353 |
on openprojects.net, I found out that while not everyone was able to get the |
| 354 |
driver to operate stably, it was working for a lot of people. |
| 355 |
</p> |
| 356 |
|
| 357 |
<p> |
| 358 |
On #nvidia, someone suggested that I make sure that the V550 wasn't sharing an |
| 359 |
IRQ with another card. Unlike the standard XFree86 driver, the accelerated |
| 360 |
NVIDIA driver requires an IRQ for proper operation. To see if it had its own |
| 361 |
dedicated IRQ, I typed <c>cat /proc/interrupts</c>, and lo and behold, my V550 |
| 362 |
was sharing an interrupt with my IDE controller. Before I explain how I solved |
| 363 |
this particular problem, I'd like to give you a brief background on IRQs. |
| 364 |
</p> |
| 365 |
|
| 366 |
<p> |
| 367 |
PCs use IRQs, and hardware interrupts in general, to allow peripheral devices, |
| 368 |
such as the video card and the disk controllers, to signal the CPU that they |
| 369 |
have data that's ready to be processed. In the old days before the PCI bus |
| 370 |
existed, it was critical that each device in the machine had its own, dedicated |
| 371 |
IRQ. In case you are still using ISA peripherals in your machine, this is still |
| 372 |
true -- all non-PCI devices should have their own dedicated IRQ. |
| 373 |
</p> |
| 374 |
|
| 375 |
</body> |
| 376 |
</section> |
| 377 |
</chapter> |
| 378 |
|
| 379 |
<chapter> |
| 380 |
<title>IRQs</title> |
| 381 |
<section> |
| 382 |
<title>IRQs and PCI</title> |
| 383 |
<body> |
| 384 |
|
| 385 |
<p> |
| 386 |
However, things are a little different with the PCI bus. PCI allocates four |
| 387 |
IRQs that can be used by the PCI/AGP cards in your system. In general, these |
| 388 |
IRQs <e>can</e> be shared among multiple devices. (If you do this, make sure |
| 389 |
that all the devices doing the sharing are PCI and AGP devices.) IRQ sharing is |
| 390 |
important, especially for modern machines that may have five PCIs and one AGP |
| 391 |
slot. Without IRQ sharing, you would be unable to have more than four IRQ-using |
| 392 |
cards in your system. |
| 393 |
</p> |
| 394 |
|
| 395 |
<p> |
| 396 |
There are, however, some limitations to PCI IRQ sharing. While modern |
| 397 |
motherboards' BIOS and Linux kernels generally support PCI IRQ sharing, certain |
| 398 |
PCI cards may simply refuse to work properly when sharing an IRQ with another |
| 399 |
device. If you're experiencing random system lockups, especially lockups that |
| 400 |
appear to be correlated with the use of a specific hardware device, you may |
| 401 |
want to try and get all your PCI devices to use their own IRQs, just to be on |
| 402 |
the safe side. The first step is to see if any devices in your system are |
| 403 |
sharing IRQs to begin with. To do this, follow these steps: |
| 404 |
</p> |
| 405 |
|
| 406 |
<ol> |
| 407 |
<li> |
| 408 |
Use the various hardware devices in your system, such as disk, sound, |
| 409 |
video, SCSI, etc. This ensures that Linux will handle interrupts for these |
| 410 |
various devices. |
| 411 |
</li> |
| 412 |
<li> |
| 413 |
<c>cat /proc/interrupts</c>, which will display a list and count of all |
| 414 |
interrupts which the Linux kernel has handled so far. Look in the far right |
| 415 |
column in this list. If two or more devices are listed in a single row, |
| 416 |
then they're sharing that particular IRQ. |
| 417 |
</li> |
| 418 |
</ol> |
| 419 |
|
| 420 |
<p> |
| 421 |
If one of the devices in question is a non-PCI device (ISA or other legacy |
| 422 |
cards) then you've found yourself an IRQ conflict, which you can attempt to fix |
| 423 |
with your BIOS, the isapnptools package, or the physical jumpers on your |
| 424 |
|
| 425 |
|
| 426 |
|
| 427 |
-- |
| 428 |
gentoo-doc-cvs@g.o mailing list |
Archives