From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <48D66DC4.8030706@domain.hid> Date: Sun, 21 Sep 2008 17:52:36 +0200 From: Gilles Chanteperdrix MIME-Version: 1.0 References: <899865CA54E4444DAF2E3639C04C5F48E4DA6A@trillian.at.omicron.at> In-Reply-To: <899865CA54E4444DAF2E3639C04C5F48E4DA6A@trillian.at.omicron.at> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Subject: Re: [Adeos-main] FW: [PATCH] repost: ARM FCSE List-Id: General discussion about Adeos List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Richard Cochran Cc: adeos-main@gna.org Richard Cochran wrote: > I posted this patch today on linux-arm-kernel, but I repeat it > here because there does not seem to be too much interest on that > list for the ARM FCSE. > > I also tried to combine this patch with ipipe for kernel 2.6.20 > running on the Intel IXDP465, but after booting I soon get a BUG. > > Anyhow, perhaps the ARM people might take a look at combining > ipipe with FCSE... Ok. Six monthes later, I finally gave a try to your patch on at91rm9200, which supports FCSE as well. When booting, I get random segmentation faults (either with or without the I-pipe), assertion which fails in glibc, and such things. I looked at the code, and have a few questions: - in cpu_switch_mm, you comment out the cache flush but keep the TLB flush, is the TLB flush really needed ? - in pmd_populate (file include/asm-arm/pgalloc.h), you populate two pmds, does not this look wrong ? I mean it looks like pmd_populate is called with the two "kinds" (translated and untranslated) addresses, which should not happen. - in exit_mmap (file mm/mmap.c), we should arrange for flush_cache_mm to flush the whole cache (I mean without consulting mm->cpu_vm_mask) to avoid keeping stale entries pointing to pages which are going to be returned to the system. I tried to change this, but I still get the random failures. - anyone still has the old FASS patch ? Perhaps by looking at it we could find what is missing in your patch. -- Gilles.