From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?ISO-8859-1?Q?Andreas_Bie=DFmann?= Date: Tue, 11 Jan 2011 14:00:50 +0100 Subject: [U-Boot] [PATCH RFC] armv7: fixloop: don't fixup if location is NULL In-Reply-To: References: <4D186A13.4080004@samsung.com> <24AA7201-FEC9-4485-837F-52CF0CBD06C4@googlemail.com> <4D284156.90305@free.fr> <20110110102052.7E30D150A44@gemini.denx.de> <20110110121453.67D51150A44@gemini.denx.de> <4D2B4002.2090408@free.fr> Message-ID: <4D2C5482.9070800@gmail.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Dear Minkyu Kang, Am 11.01.2011 11:57, schrieb Minkyu Kang: >> Regardless of the patch, if your code writes to panel_info or any other BSS >> variable before relocation it will trash the relocation tables that exist at >> BSS location at this point. >> >> IOW, accessing BSS before relocation is forbidden, not just out of fancy, >> but for a serious reason. > > This patch is not for accessing BSS before relocation, > it's for prevent exceptions. The real error is writing to BSS before relocation. This leads to a corrupted .rel.dyn section which is placed at the same address as .bss at this moment (bss is overloaded to save space). If you look in your ELF (e.g. readelf -R .rel.dyn u-boot) you may see, that the .rel.dyn section does _not_ include a pointer to 0x0 with relative relocation (0x17) as you showed in a previous post. If you look in your u-boot.map you may find the function in question (test_func() was it in your example) is placed in .bss section. Setting the function pointer to 0 (e.g. test_func() = NULL, as described in previous mail) before relocation will destroy your .rel.dyn section and then you will see a zero in .rel.dyn section at some place ... please investigate the ELF and do not step through the code to find those issues. I may be wrong, please show it to us. regards Andreas Bie?mann