CINXE.COM

<?xml version="1.0" encoding="UTF-8" standalone="yes"?> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>LKML: Ladislav Michl: Re: [PATCH] ds1337 4/4</title><link href="/css/message.css" rel="stylesheet" type="text/css" /><link href="/css/wrap.css" rel="alternate stylesheet" type="text/css" title="wrap" /><link href="/css/nowrap.css" rel="stylesheet" type="text/css" title="nowrap" /><link href="/favicon.ico" rel="shortcut icon" /><script src="/js/simple-calendar.js" type="text/javascript"></script><script src="/js/styleswitcher.js" type="text/javascript"></script><link rel="alternate" type="application/rss+xml" title="lkml.org : last 100 messages" href="/rss.php" /><link rel="alternate" type="application/rss+xml" title="lkml.org : last messages by Ladislav Michl" href="/groupie.php?aid=21763" /><script> var _paq = window._paq = window._paq || []; /* tracker methods like "setCustomDimension" should be called before "trackPageView" */ _paq.push(["setDoNotTrack", true]); _paq.push(["disableCookies"]); _paq.push(['trackPageView']); _paq.push(['enableLinkTracking']); (function() { var u="//m.lkml.org/"; _paq.push(['setTrackerUrl', u+'matomo.php']); _paq.push(['setSiteId', '1']); var d=document, g=d.createElement('script'), s=d.getElementsByTagName('script')[0]; g.async=true; g.src=u+'matomo.js'; s.parentNode.insertBefore(g,s); })(); </script></head><body onload="es.jasper.simpleCalendar.init();" itemscope="itemscope" itemtype="http://schema.org/BlogPosting"><table border="0" cellpadding="0" cellspacing="0"><tr><td width="180" align="center"><a href="/"><img style="border:0;width:135px;height:32px" src="/images/toprowlk.gif" alt="lkml.org" /></a></td><td width="32">聽</td><td class="nb"><div><a class="nb" href="/lkml"> [lkml]</a> 聽 <a class="nb" href="/lkml/2005"> [2005]</a> 聽 <a class="nb" href="/lkml/2005/4"> [Apr]</a> 聽 <a class="nb" href="/lkml/2005/4/8"> [8]</a> 聽 <a class="nb" href="/lkml/last100"> [last100]</a> 聽 <a href="/rss.php"><img src="/images/rss-or.gif" border="0" alt="RSS Feed" /></a></div><div>Views: <a href="#" class="nowrap" onclick="setActiveStyleSheet('wrap');return false;">[wrap]</a><a href="#" class="wrap" onclick="setActiveStyleSheet('nowrap');return false;">[no wrap]</a> 聽 <a class="nb" href="/lkml/mheaders/2005/4/8/124" onclick="this.href='/lkml/headers'+'/2005/4/8/124';">[headers]</a>聽 <a href="/lkml/bounce/2005/4/8/124">[forward]</a>聽 </div></td><td width="32">聽</td></tr><tr><td valign="top"><div class="es-jasper-simpleCalendar" baseurl="/lkml/"></div><div class="threadlist">Messages in this thread</div><ul class="threadlist"><li class="root"><a href="/lkml/2005/3/31/279">First message in thread</a></li><li><a href="/lkml/2005/4/7/271">Ladislav Michl</a><ul><li><a href="/lkml/2005/4/8/104">"Jean Delvare"</a><ul><li class="origin"><a href="/lkml/2005/4/8/182">Ladislav Michl</a><ul><li><a href="/lkml/2005/4/8/182">"Jean Delvare"</a></li><li><a href="/lkml/2005/4/8/206">James Chapman</a><ul><li><a href="/lkml/2005/4/10/130">Ladislav Michl</a></li></ul></li></ul></li></ul></li></ul></li></ul></td><td width="32" rowspan="2" class="c" valign="top"><img src="/images/icornerl.gif" width="32" height="32" alt="/" /></td><td class="c" rowspan="2" valign="top" style="padding-top: 1em"><table><tr><td><table><tr><td class="lp">Date</td><td class="rp" itemprop="datePublished">Fri, 8 Apr 2005 14:35:45 +0200</td></tr><tr><td class="lp">Subject</td><td class="rp" itemprop="name">Re: [PATCH] ds1337 4/4</td></tr><tr><td class="lp">From</td><td class="rp" itemprop="author">Ladislav Michl <></td></tr></table></td><td></td></tr></table><pre itemprop="articleBody">On Fri, Apr 08, 2005 at 01:08:46PM +0200, Jean Delvare wrote: > > Add support for DS1339. The only difference against DS1337 is Trickle > > Charge register at address 10h, which is used to enable battery or gold > > cap charging. Please note that value may vary for different batteries, > > so it should be made module parameter. 0xaa is sane default and also > > matches my board ;-) > > "Sane default" is a non-sense here. A sane default is that loading a > real-time clock driver should not affect the battery at all IMHO. > > Can you tell us more on that battery thing? I admit I don't exatcly get > what it is. Which type of battery are we talking about? Are there > similar drivers in the kernel tree already? Sorry, I have no clue what devices are using it and what may come to board designer's mind. This chip will be most likely used in embedded, where every sane developer is expected to review drivers he is using. Also in such situation driver is compiled staticaly. (And in ideal world firmware (such as U-Boot, RedBoot, etc.) should set this register). > Sounds weird to me that loading a driver would enable charging of a > battery, and removing it wouldn't disable it. And since the driver > might not be removed, it would possibly make more sense to have an entry > in /sys to enable and disable this thing. Disabling battery charge makes no sense. The only reason I can think about is when suspending device, so it is likely pm job. /sys entry might help as well, but I do not see any point making driver more complicated and bigger just to make someone else happy. Golden rule is: implement features as needed :) > Also, arbitrarily picking one of the 6 possible charging modes just > because it matches your board is a bad idea. It looks like a value which > should be set on a per-board basis, rather than picked randomly by the > user, so as to avoid accidental hardware breakage. Well free to provide that way, so far I'm the only user so I did what is usefull for me [*]. Everyone is welcome to change it to more generic way. > > +static int ds1339_charge = 0xaa; > > I see little reason why this would be a global variable rather than a > define (let alone the fact that this shouldn't be hardcoded at all > IMHO, as I just explained). [*] This way it can be easily turned into module parameter or whatever someone who will need different value may want to do. Lets choose define, later it can be converted to module parameter. Gcc is smart enough to optimize charge enable code away. #define ds1339_charge 0 > > +/* > > + * DS1339 has Trickle Charge register at address 10h. During a multibyte > > + * access, when the address pointer reaches the end of the register space, > > + * it wraps around to location 00h. > > + * We read 17 bytes from device and compare first and last one. If they > > + * are same it is most likely DS1337 chip. > > + */ > > +static int ds1337_is_ds1339(struct i2c_client *client) > > +{ > > + char buf[17], addr = 0; > > + struct i2c_msg msg[2]; > > + int result; > > + > > + msg[0].addr = client->addr; > > + msg[0].flags = 0; > > + msg[0].len = 1; > > + msg[0].buf = &addr; > > + > > + msg[1].addr = client->addr; > > + msg[1].flags = I2C_M_RD; > > + msg[1].len = sizeof(buf); > > + msg[1].buf = buf; > > + > > + result = i2c_transfer(client->adapter, msg, 2); > > + if (result < 0) { > > + dev_err(&client->dev, "error reading data! %d\n", result); > > + return 0; > > + } else > > + return (buf[0] == buf[16]) ? 0 : 1; > > +} > > This will fail eventually. The first register is the seconds count, which > by definition is changing. I2C is slow, by the time you wrap over the > register range, the value could have changed. The datasheet explicitely > says that the register cache will refresh on address wrapping. I was running test overnight and didn't meet any single case when this happen. Perhaps device also needs to see start condition? > Also, 0x00 is a possible value for both the seconds count and the battery > register, so you could miss a DS1339 at times. > > One possible check to start with would be on the value of the additional > register itself. It has only 7 possible values. (...) Eh? Register is 8bit, that's 256 combinations. > (...) But of course it would be better if we could default to a DS1337 > and find a way to identify the DS1339, rather than the (unsafe) other > way around. I also don't know what a DS1337 will answer for this > missing register. Maybe James can help? > > One possibility would be to start reading at 0x0E instead of 0x00, > because register 0x00 is the control register and is the only one which > will not change in our back as far as I can see. Oh, and the additional > battery register too. But this still doesn't sound like a bulletproof > method to me (we depend on the seconds and possibly minutes count > again). So it would be better to additionally perform the same tests > that were done on the non-wrapped registers for the regular DS1337 > detection, but on the wrapped area. > > The problem here is that all this will significantly increase the > detection delay. That's probably overkill, see above. Look, the only difference between ds1339 and ds1337 is Trickle Charge register. We won't touch it by default and if anyone wants to use it, he need to provide its value. In that case he also knows it is DS1339 and also knows what battery is wired, so he knows charging current. I hope it is also clear that without picking one "sane default" there is no point to do any detection at all. > Yet another method would be to write a non-significant value to the > battery register, such as 0x80. If we can read it back then it has to be > a DS1339. But what effect will it have on the DS1337? I'd hope none, > but this better be verified. And in general I don't much like using > register writes in detection methods. You will change register 00h. > Could you please provide the output of i2cdump in b and c modes for your > DS1339 chip? This might help finding an detection method. A similar dump > for the DS1337 would help too. Sorry it seems there is no easy way to cross-compile this package. I have admit I haven't tried too hard. > > - const char *name = ""; > > + const char *name; > (...) > > + default: > > + name = ""; > > + } > > This is noise. No it is not. Look at generated machine code for each case and try to do it for at least two architectures different from i386 ;-) I'm bored with arguing about such changes, do what you think is right, but it will remain in patch. The rest is up to you. Also note that this very way of rejecting tiny cleanups leads to hardly readable and unoptimized code and after some greater amount of time code rewrite occurs anyway (once someone is sufficienly upset). Have a nice weekend, ladis - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at <a href="http://vger.kernel.org/majordomo-info.html">http://vger.kernel.org/majordomo-info.html</a> Please read the FAQ at <a href="http://www.tux.org/lkml/">http://www.tux.org/lkml/</a> </pre></td><td width="32" rowspan="2" class="c" valign="top"><img src="/images/icornerr.gif" width="32" height="32" alt="\" /></td></tr><tr><td align="right" valign="bottom"> 聽 </td></tr><tr><td align="right" valign="bottom">聽</td><td class="c" valign="bottom" style="padding-bottom: 0px"><img src="/images/bcornerl.gif" width="32" height="32" alt="\" /></td><td class="c">聽</td><td class="c" valign="bottom" style="padding-bottom: 0px"><img src="/images/bcornerr.gif" width="32" height="32" alt="/" /></td></tr><tr><td align="right" valign="top" colspan="2"> 聽 </td><td class="lm">Last update: 2005-04-08 14:39 聽聽 [from the cache] 漏2003-2020 <a href="http://blog.jasper.es/">Jasper Spaans</a>|hosted at <a href="https://www.digitalocean.com/?refcode=9a8e99d24cf9">Digital Ocean</a> and my Meterkast|<a href="http://blog.jasper.es/categories.html#lkml-ref">Read the blog</a></td><td>聽</td></tr></table><script language="javascript" src="/js/styleswitcher.js" type="text/javascript"></script></body></html>