FreeCalypso > hg > fc-magnetite

view blobs/patches/frame_na7_db_ir-systrace.patch @ 516:1ed9de6c90bd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
src/g23m-gsm/sms/sms_for.c: bogus malloc removed The new error handling code that was not present in TCS211 blob version contains a malloc call that is bogus for 3 reasons: 1) The memory allocation in question is not needed in the first place; 2) libc malloc is used instead of one of the firmware's proper ways; 3) The memory allocation is made inside a function and then never freed, i.e., a memory leak. This bug was caught in gcc-built FreeCalypso fw projects (Citrine and Selenite) because our gcc environment does not allow any use of libc malloc (any reference to malloc produces a link failure), but this code from TCS3.2 is wrong even for Magnetite: if this code path is executed repeatedly over a long time, the many small allocations made by this malloc call without a subsequent free will eventually exhaust the malloc heap provided by the TMS470 environment, malloc will start returning NULL, and the bogus code will treat it as an error. Because the memory allocation in question is not needed at all, the fix entails simply removing it.
author Mychaela Falconia <falcon@freecalypso.org>
date Sun, 22 Jul 2018 06:04:49 +0000
parents acb07ce22054
children
line wrap: on
line source

# TI's original version of the pf_TaskEntry() function in the guts of GPF
# does this number once all GPF tasks have entered their main loop:
#
#    TraceMask[0] = 0;
#
# The effect of this line is that "system" trace output gets disabled;
# in particular, "system primitive" requests like querying for existing
# TRACECLASS of various entities or for the GPF version elicit no response
# whatsoever until one sends a TRACECLASS setting sysprim which restores
# the TC_SYSTEM bit in TraceMask[0].
#
# In FreeCalypso we consider this clearing of TraceMask[0] to be either
# a bug or a misfeature, and we would like to change it.  In the case of
# the GPF code in question, we do have a copy of frame.c source (the module
# where the offending pf_TaskEntry() function is implemented) which we believe
# is the same code as what's in the TCS211 version of GPF, but TCS211 is not
# set up to recompile any part of GPF from source, and there are other modules
# in the same frame_na7_db_ir.lib for which we don't have the original source
# (os_*), hence it is easier to do a binary lib patch.
#
# The present patch is a binary form of:
#
# -TraceMask[0] = 0;
# +TraceMask[0] = TC_SYSTEM;

[frame.obj]

.text 2ee 10