LKML: Pekka Enberg: Re: [PATCH 01/13] [RFC] ipath basic headers

<?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: Pekka Enberg: Re: [PATCH 01/13] [RFC] ipath basic headers</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 Pekka Enberg" href="/groupie.php?aid=23443" /><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/2005/12"> [Dec]</a> 聽 <a class="nb" href="/lkml/2005/12/17"> [17]</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/2005/12/17/23" onclick="this.href='/lkml/headers'+'/2005/12/17/23';">[headers]</a>聽 <a href="/lkml/bounce/2005/12/17/23">[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/2005/12/16/290">First message in thread</a></li><li><a href="/lkml/2005/12/16/290">Roland Dreier</a><ul><li><a href="/lkml/2005/12/16/293">Roland Dreier</a><ul><li><a href="/lkml/2005/12/16/289">Roland Dreier</a><ul><li><a href="/lkml/2005/12/16/291">Roland Dreier</a><ul><li><a href="/lkml/2005/12/16/303">Roland Dreier</a></li><li><a href="/lkml/2005/12/17/25">Pekka Enberg</a></li><li><a href="/lkml/2005/12/17/31">Christoph Hellwig</a></li><li><a href="/lkml/2005/12/17/78">Andrew Morton</a></li></ul></li></ul></li><li class="origin"><a href="/lkml/2005/12/17/97">Pekka Enberg</a><ul><li><a href="/lkml/2005/12/17/97">Robert Walsh</a></li></ul></li><li><a href="/lkml/2005/12/17/29">Christoph Hellwig</a><ul><li><a href="/lkml/2005/12/17/96">(Eric W. Biederman)</a><ul><li><a href="/lkml/2005/12/17/129">Andi Kleen</a></li></ul></li><li><a href="/lkml/2005/12/17/98">Robert Walsh</a><ul><li><a href="/lkml/2005/12/17/100">Arjan van de Ven</a></li></ul></li></ul></li><li><a href="/lkml/2005/12/17/77">Andrew Morton</a><ul><li><a href="/lkml/2005/12/17/102">Robert Walsh</a><ul><li><a href="/lkml/2005/12/17/126">Andrew Morton</a></li></ul></li><li><a href="/lkml/2005/12/19/207">Robert Walsh</a></li></ul></li></ul></li><li><a href="/lkml/2005/12/17/30">Christoph Hellwig</a><ul><li><a href="/lkml/2005/12/17/37">Roland Dreier</a></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">Sat, 17 Dec 2005 14:33:16 +0200</td></tr><tr><td class="lp">From</td><td class="rp" itemprop="author">Pekka Enberg <></td></tr><tr><td class="lp">Subject</td><td class="rp" itemprop="name">Re: [PATCH 01/13] [RFC] ipath basic headers</td></tr></table></td><td></td></tr></table><pre itemprop="articleBody">Hi Roland, On 12/17/05, Roland Dreier <rolandd@cisco.com> wrote: > +/* > + * This file contains defines, structures, etc. that are used > + * to communicate between kernel and user code. > + */ > + > +#ifdef __KERNEL__ > +#include <linux/ioctl.h> > +#include <linux/uio.h> > +#include <asm/atomic.h> > +#else /* !__KERNEL__; user mode */ > +#include <sys/ioctl.h> > +#include <sys/uio.h> > +#include <sys/types.h> > +#include <stdint.h> > + > +/* these aren't implemented for user mode, which is OK until we multi-thread */ > +typedef struct _atomic { > + uint32_t counter; > +} atomic_t; /* no atomic_t type in user-land */ > +#define atomic_set(a,v) ((a)->counter = (v)) > +#define atomic_inc_return(a) (++(a)->counter) > +#define likely(x) (x) > +#define unlikely(x) (x) > + > +#define yield() sched_yield() > + > +/* > + * too horrible to try and use the kernel get_cycles() or equivalent, > + * so define and inline it here > + */ > + > +#if !defined(rdtscll) > +#if defined(__x86_64) || defined(__i386) > +#define rdtscll(v) do {uint32_t a,d;asm volatile("rdtsc" : "=a" (a), "=d" (d)); \ > + (v) = ((uint64_t)a) | (((uint64_t)d)<<32); \ > +} while(0) > +#else > +#error "No cycle counter routine implemented yet for this platform" > +#endif > +#endif /* !defined(rdtscll) */ Do we really need this ugly userspace emulation code in the kernel? > +/* > + * this is used for very short copies, usually 1 - 8 bytes, > + * *NEVER* to the PIO buffers!!!!!!! use ipath_dwordcpy for longer > + * copies, or any copy to the PIO buffers. Works for 32 and 64 bit > + * gcc and pathcc > + */ > +static __inline__ void ipath_shortcopy(void *dest, void *src, uint32_t cnt) > +{ > + void *ssv, *dsv; > + uint32_t csv; > + __asm__ __volatile__("cld\n\trep\n\tmovsb":"=&c"(csv), "=&D"(dsv), > + "=&S"(ssv) > + :"0"(cnt), "1"(dest), "2"(src) > + :"memory"); > +} > + > +/* > + * optimized word copy; good for rev C and later opterons. Among the best for > + * short copies, and does as well or slightly better than the optimizization > + * guide copies 6 and 8 at 2KB. > + */ > +void ipath_dwordcpy(uint32_t * dest, uint32_t * src, uint32_t ndwords); What is this used for? Why can't yo use memcpy? > +#define round_up(v,sz) (((v) + (sz)-1) & ~((sz)-1)) Please use ALIGN(). > +/* These are used in the driver, don't use them elsewhere */ > +#define _IPATH_SIMFUNC_IOCTL_LOW 1 > +#define _IPATH_SIMFUNC_IOCTL_HIGH 7 > + > +/* > + * These tell the driver which ioctl's belong to the diags interface. > + * As above, don't use them elsewhere. > + */ > +#define _IPATH_DIAG_IOCTL_LOW 100 > +#define _IPATH_DIAG_IOCTL_HIGH 109 [snip, snip] You seem to be introducing loads of new ioctls. Any reason you can't use sysfs and/or configfs? > +/* macros for processing rcvhdrq entries */ > +#define ips_get_hdr_err_flags(StartOfBuffer) *(((uint32_t *)(StartOfBuffer))+1) > +#define ips_get_index(StartOfBuffer) (((*((uint32_t *)(StartOfBuffer))) >> \ > + INFINIPATH_RHF_EGRINDEX_SHIFT) & INFINIPATH_RHF_EGRINDEX_MASK) > +#define ips_get_rcv_type(StartOfBuffer) ((*(((uint32_t *)(StartOfBuffer))) >> \ > + INFINIPATH_RHF_RCVTYPE_SHIFT) & INFINIPATH_RHF_RCVTYPE_MASK) > +#define ips_get_length_in_bytes(StartOfBuffer) \ > + (uint32_t)(((*(((uint32_t *)(StartOfBuffer))) >> \ > + INFINIPATH_RHF_LENGTH_SHIFT) & INFINIPATH_RHF_LENGTH_MASK) << 2) > +#define ips_get_first_protocol_header(StartOfBuffer) (void *) \ > + ((uint32_t *)(StartOfBuffer) + 2) > +#define ips_get_ips_header(StartOfBuffer) ((ips_message_header_typ *) \ > + ((uint32_t *)(StartOfBuffer) + 2)) > +#define ips_get_ipath_ver(ipath_header) (((ipath_header) >> INFINIPATH_I_VERS_SHIFT) \ > + & INFINIPATH_I_VERS_MASK) Please use static inlines instead for readability. - 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: 2005-12-17 13:36 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>

CINXE.COM

LKML: Pekka Enberg: Re: [PATCH 01/13] [RFC] ipath basic headers