From: Simon Goldschmidt Date: Wed, 17 Jul 2019 20:18:23 +0200 Subject: [U-Boot] [PATCH] spl: implement stack usage check In-Reply-To: <20190717180722.GF20116@bill-the-cat> References: <20190716203036.24430-1-simon.k.r.goldschmidt@gmail.com> <20190717180722.GF20116@bill-the-cat> Message-ID: <09a758ba-d7e8-a3e2-4e29-6788d5cd929a@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 Am 17.07.2019 um 20:07 schrieb Tom Rini: > On Tue, Jul 16, 2019 at 10:30:36PM +0200, Simon Goldschmidt wrote: > >> This implements a stack usage check in SPL. >> >> Many boards start up SPL to run code + data from one common, rather small >> SRAM. To implement a sophisticated SPL binary size limit on such boards, >> the stack size (as well as malloc size and global data size) must be >> subtracted from this SRAM size. >> >> However, to do that properly, the stack size first needs to be known. >> >> This patch adds a new Kconfig option: >> - SPL_SYS_REPORT_STACK_F_USAGE: memset(0xaa) the whole area of the stack >> very early and check stack usage based on this constant later before the >> stack is switched to DRAM >> >> Initializing the stack and checking it is implemented in weak functions, >> in case a board does not use the stack as saved in gd->start_addr_sp. >> >> Signed-off-by: Simon Goldschmidt > > Very nice! I wish I had thought of something like that years ago when > Wolfgang asked and I came up with the -fstack-usage stuff we have today. Well, I think theoretically, a compile-time approach is better than what I did. However, that -fstack-usage thing did not work out for me (I'd have to look into that again), so I implemented this runtime check. On the positive side, pre-relocation stack usage doesn't change too much, so the old rule of thumb of checking it once and the taking roughly the double as config option should be good ;-) Regards, Simon