From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andreas Glatz In-Reply-To: <1250696784.3014.31.camel@domain.hid> References: <1250696784.3014.31.camel@domain.hid> Content-Type: text/plain Content-Transfer-Encoding: 7bit Date: Fri, 21 Aug 2009 11:44:47 -0400 Message-Id: <1250869487.4031.2.camel@domain.hid> Mime-Version: 1.0 Subject: Re: [Xenomai-core] [Xenomai-help] Xenomai v2.4.9 List-Id: Xenomai life and development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Philippe Gerum Cc: xenomai@xenomai.org Hi, Is there a reason why the following patch didn't make it into 2.4.9? See: https://mail.gna.org/public/xenomai-help/2009-08/msg00023.html I included it into our system. I didn't see any problems so far... Regards, Andreas On Wed, 2009-08-19 at 17:46 +0200, Philippe Gerum wrote: > Here is the ninth maintenance release for the v2.4.x branch: > > http://download.gna.org/xenomai/stable/xenomai-2.4.9.tar.bz2 > > Special note to PowerPC (32/64) users: the FPU management code has been > sanitized an awful lot in this release compared to 2.4.8 and earlier > for this architecture. Really. Several bugs which may badly break the > floating-point registers state "in some circumstances" have > been fixed. > > Therefore I would strongly recommend to upgrade your Xenomai baseline > to 2.4.9, and ideally, kernel 2.6/powerpc users should also move their > I-pipe support to 2.6.30-2.7-00, or at the very least backport _both_ of > the following patches, in that order, to their kernel tree: > > 1. http://git.denx.de/?p=ipipe-2.6.git;a=blobdiff;f=arch/powerpc/kernel/fpu.S;h=503de51bed94cc2fc4d17b4342885d0329ba5c1b;hp=2436df33c6f4433d7e7f54b88caad6ba570cf3f6;hb=4a9cffba3407f1238646847aad06712d8952a310;hpb=4d121ab41bda8a269c20b0a1077af8f3ef400105 > 2. http://git.denx.de/?p=ipipe-2.6.git;a=commitdiff;h=0024416dcce817854cbf80eff420b21ef615672c;hp=d0353a6d1b4b4af7f6990dfa27055c799b3756a9 > > People running a 2.4.25 kernel (the only one officially supported by > Xenomai from the legacy 2.4/ppc kernel series) should definitely > upgrade to the following I-pipe patch, to close a few important > issues, including the FPU ones: > http://download.gna.org/adeos/patches/v2.4/ppc/adeos-ipipe-2.4.25-ppc-DENX-2.2-01.patch > > Panic mode mitigation: This is a PowerPC-related issue only. Other > architectures were not affected by those bugs. > > = You have been warned = > > Not scared enough? Ok. Here is the short log for v2.4.9: > > Gilles Chanteperdrix (3): > arm: fix syscalls with gcc 4.3.2 > arm: Disable preemptible switch. > posix: fix shm leak > > Graf Yang (1): > blackfin: Update __sched_latency for blackfin processors > > Matteo Facchinetti (1): > rtcan: fix MPC5xxx_GPIO definition for 2.6.2[0-4] kernels > > Philippe Gerum (54): > powerpc: add calibration data for Freescale's ADS8272 > powerpc: add calibration data for TQ's TQM8560 > powerpc: upgrade Adeos/powerpc support > build: upgrade to autoconf >= 2.62 > powerpc: add calibration info for 440EP Yosemite > powerpc: update Adeos/powerpc support > powerpc: unify calibration values for PA-Semi > powerpc: upgrade I-pipe/2.4 support to 2.4.25-ppc-2.0-01 > x86: upgrade I-pipe support to 2.6.29.5-x86-2.4-01 > x86: upgrade I-pipe to 2.6.29.5-x86-2.4-02 > powerpc: init FPU for userland Xenomai threads as well > powerpc: discard useless backup field > powerpc: fix fpu backup area layout > nucleus: fix potential init race of the main pod > powerpc: sanitize kthread init stack setup > nucleus: port to 2.6.30 > native: force no more than PTHREAD_STACK_MIN bytes for stack size > powerpc: enable FE0/FE1 exceptions during fpu init > powerpc: flip FE0/FE1 when saving/restoring the FPU in RT mode > x86: fix wrap_switch_iobitmap() for 2.6.30 > x86: remove now useless I/O bitmap switch code > nucleus: do not dereference thread memory after deletion > powerpc: raise SYS_STACKPOOLSZ to 128Kb for 64bit archs > nucleus: use larger default sizes for heap and stack pool > powerpc: reserve stack overhead in kthread trampoline > powerpc: sanitize FPU handling helpers > testsuite: add missing compiler barriers to switchtest > testsuite: do not specify stack sizes lower than PTHREAD_STACK_MIN > powerpc: upgrade legacy I-pipe support to 2.4.25-ppc-2.1-00 > powerpc: upgrade legacy I-pipe support to 2.6.20-powerpc-1.9-00 > powerpc: upgrade I-pipe support to 2.6.30-powerpc-2.6-03 > arm: upgrade I-pipe support to 2.6.28-arm-1.12-03 > arm: upgrade I-pipe support to 2.6.28-arm-1.13-01 > x86: fix 2.6.30 build in 32bit mode > x86: upgrade I-pipe support to 2.6.30-x86-2.4-05 > powerpc: upgrade I-pipe support to 2.6.30-powerpc-2.7-00 > rtai: fix build with CONFIG_PROC_FS disabled > blackfin: upgrade I-pipe support to 2.6.30-1.11-00 > asm-generic: stop using deprecated find_task_by_pid_type_ns() > x86: fix include for kernel >= 2.6.30 > doc: fix GIT tree detection > blackfin: do not consider IPEND[4] for scheduling deferral > powerpc: reinstate FE0/FE1 depending on the per-task fpexc_mode > powerpc: do not leak MSR_FP when back to secondary mode > powerpc: upgrade legacy I-pipe support to 2.4.25-ppc-2.2-01 > nucleus: trap wrong interrupt state upon return from hardening > nucleus: postpone sanity check to allow for signal receipt > arm: upgrade I-pipe support to 2.6.30-arm-1.14-00, 2.6.28-arm-1.12-04, 2.6.29-arm-1.13-02 > scripts: use default autoconf version > powerpc: upgrade legacy I-pipe support to 2.6.20-powerpc-1.9-01 > build: update version stamp > build: bootstrap > doc: regenerate documentation > > Stefan Kisdaroczi (1): > debian: fix kernel patch preparation script not to truncate init/Kconfig >