From mboxrd@z Thu Jan 1 00:00:00 1970 From: Nathan_Lynch@mentor.com (Nathan Lynch) Date: Tue, 18 Mar 2014 12:17:39 -0500 Subject: [PATCH v4] ARM: vDSO gettimeofday using generic timer architecture In-Reply-To: References: <1394734769-32760-1-git-send-email-nathan_lynch@mentor.com> Message-ID: <53287FB3.1050103@mentor.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org Hi David, On 03/17/2014 07:49 PM, David Riley wrote: > Hi Nathan, > > I gave this a try against a 3.10 system and ran into issues when > vdso.so grew to two pages because of of a large .GCC.command.line > section. __get_datapage was returning an address that was at the > beginning of the second code page instead of the data page since it > didn't account for the additional section. I got it working by adding > .GCC.command.line to the DISCARD group in the linker script, but in > general it feels as if this code is a bit fragile since it depends on > knowing exactly which sections exist in the .so for the _vdso_data > symbol to be correct. Thanks for giving this a try and reporting the issue. I'm guessing you use -frecord-gcc-switches. I agree this aspect of the vDSO could be more robust -- at least the build should be made to fail when wrong code is generated. I'll look into it, and I'd appreciate any suggestions or additional review, especially of the linker script.