From: Jovi Z. <boo...@gm...> - 2012-10-19 08:28:37
|
Hi, I encounter a problem in strace 4.5.19 version(for arm), strace detach as soon as attach. I got below old bug report in this mailing list, which have same problem with my situation. http://www.mail-archive.com/str...@li.../msg01669.html Unfortunately, I don't get answer on that bug report, so is there anyone can tell me that this bug is already fixed in strace or kernel? Thanks! [bug] strace detach as soon as attach OneGun Lee Wed, 07 Jul 2010 23:13:48 -0700 Hi, all I tried to use strace-4.5.15(and .20) on linux-2.6.29.4, arm-cortex-a8. but strace was detached as soon as was attached one process. # strace -p 1711 -d Process 1711 attached - interrupt to quit [wait(0x137f) = 1711] pid 1711 stopped, [SIGSTOP] [wait(0x57f) = 1711] pid 1711 stopped, [SIGTRAP] Process 1711 detached So, I debuged strace. After attach, strace try to ptrace(PTRACE_PEEKTEXT) and input param is kernel space's address(0xFFFFxxxx). This kernel space address's data is scno(restart_syscall). address = 0xffff051c, data = 0xef900000(restart_syscall) This command is sent kernel mode. and than, arch_ptrace,kernel function, try to get data this address. so this function call access_process_vm() -> get_user_pages(). get_user_pages() function make error because start address param is kernel space's address. so strace is detached. I think that it is bug to need fix. If address to want to get data is in kernel space, access_process_vm() just read data. don't call get_user_pages(). Or strace have not try to peek data, if address is in kernel space. How do you think about this problem? |
From: Jovi Z. <boo...@gm...> - 2012-10-19 09:18:10
|
On Fri, Oct 19, 2012 at 4:28 PM, Jovi Zhang <boo...@gm...> wrote: > Hi, > > I encounter a problem in strace 4.5.19 version(for arm), strace detach as > soon as attach. > > I got below old bug report in this mailing list, which have same problem > with my situation. > > > http://www.mail-archive.com/str...@li.../msg01669.html > > Unfortunately, I don't get answer on that bug report, so is there anyone > can tell me that this bug is already fixed in strace or kernel? > Thanks! > > [bug] strace detach as soon as attach > > OneGun Lee > Wed, 07 Jul 2010 23:13:48 -0700 > > Hi, all > > I tried to use strace-4.5.15(and .20) on linux-2.6.29.4, arm-cortex-a8. > > but strace was detached as soon as was attached one process. > > # strace -p 1711 -d > Process 1711 attached - interrupt to quit > [wait(0x137f) = 1711] > pid 1711 stopped, [SIGSTOP] > [wait(0x57f) = 1711] > pid 1711 stopped, [SIGTRAP] > Process 1711 detached > > So, I debuged strace. > After attach, strace try to ptrace(PTRACE_PEEKTEXT) and input param > is kernel space's address(0xFFFFxxxx). > This kernel space address's data is scno(restart_syscall). > > address = 0xffff051c, data = 0xef900000(restart_syscall) > > This command is sent kernel mode. > and than, arch_ptrace,kernel function, try to get data this address. > so this function call access_process_vm() -> get_user_pages(). > > get_user_pages() function make error because start address param is kernel > space's address. > so strace is detached. > I think that it is bug to need fix. > If address to want to get data is in kernel space, access_process_vm() just > read data. don't call get_user_pages(). > Or > strace have not try to peek data, if address is in kernel space. > > How do you think about this problem? > > It seems that there had one patch target for fix this issue, but I checked git history of strace, I don't find this patch merged, so it means that this issue still on there. http://marc.info/?l=strace&m=128703091611617 |