Re: [Crash-utility] [PATCH] Do not pass through the 'sy' command to GDB

在 2021年03月23日 14:11, HAGIO KAZUHITO(萩尾　一仁) 写道: > Hi Lianbo, > > -----Original Message----- >> The 'sy' command may be misused, which is mistakenly considered as >> the 'symbol-file' command in the crash utility, this will discard >> symbol table from the current symbol file, and eventually caused the >> failure of crash utility after executing the 'sys' command as below: >> >> crash> sy >> GNU_GET_DATATYPE[sy]: returned via gdb_error_hook >> Discard symbol table from `/usr/lib/debug/usr/lib/modules/5.11.0-2.el9.x86_64/vmlinux'? (y or n) Please >> answer y or n. >> Discard symbol table from `/usr/lib/debug/usr/lib/modules/5.11.0-2.el9.x86_64/vmlinux'? (y or n) No symbol >> file now. >> crash> sys >> GNU_GET_SYMBOL_TYPE: returned via gdb_error_hook >> double free or corruption (!prev) >> Aborted (core dumped) >> >> Actually, the 'symbol-file' command has been added to the gdb-prohibited >> list. To prevent current error, let's add the 'sy' command to the list >> so that the crash utility does not pass the 'sy' command directly to GDB. >> >> Signed-off-by: Lianbo Jiang <lijiang(a)redhat.com&gt; > > ok, thanks. > > Acked-by: Kazuhito Hagio <k-hagio-ab(a)nec.com&gt; > Thank you for the comment, Kazu. > I'm afraid that this kind of patch work can be endless, as gdb accepts > abbreviations of its commands. but this is the first one I see and > likely to hit it accidentally, I ack this patch. > Agree with you. It can not cover all cases that users intentionally input the 'wrong' commands, but it could be necessary to prevent the critical error by the accidental input. > However, if similar issues are reported, it would be good to think about > a better way. (e.g. prefix matching? I'm not sure if it's feasible and > they're worth taking efforts to fix..) > The command name abbreviations are useful in the GDB, but for the crash-utility, need to carefully cope with this feature. It could not be worth covering all cases about this issue as we mentioned above. So far, I haven't found a good way to fix these problems. Maybe we could provide a document to clarify such issues?