lkml.org 
[lkml]   [2011]   [Oct]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: [PATCH v5 3.1.0-rc4-tip 12/26] Uprobes: Handle breakpoint and Singlestep
    On 09/20, Srikar Dronamraju wrote:
    >
    > @@ -1285,6 +1286,9 @@ static struct task_struct *copy_process(unsigned long clone_flags,
    > INIT_LIST_HEAD(&p->pi_state_list);
    > p->pi_state_cache = NULL;
    > #endif
    > +#ifdef CONFIG_UPROBES
    > + p->utask = NULL;
    > +#endif

    I am not sure I understand this all right, but I am not sure this
    is enough...

    What if the forking task (current) is in UTASK_BP_HIT state?

    IOW, uprobe replaces the original syscall insn with "int3", then we
    enter the kernel from the xol_vma. The new child has the same
    modified instruction pointer (pointing to nowhere without CLONE_VM)
    and in any case it doesn't have TIF_SINGLESTEP.

    No?

    > +void uprobe_notify_resume(struct pt_regs *regs)
    > +{
    > + struct vm_area_struct *vma;
    > + struct uprobe_task *utask;
    > + struct mm_struct *mm;
    > + struct uprobe *u = NULL;
    > + unsigned long probept;
    > +
    > + utask = current->utask;
    > + mm = current->mm;
    > + if (!utask || utask->state == UTASK_BP_HIT) {
    > + probept = get_uprobe_bkpt_addr(regs);
    > + down_read(&mm->mmap_sem);
    > + vma = find_vma(mm, probept);
    > + if (vma && valid_vma(vma))
    > + u = find_uprobe(vma->vm_file->f_mapping->host,
    > + probept - vma->vm_start +
    > + (vma->vm_pgoff << PAGE_SHIFT));
    > + up_read(&mm->mmap_sem);
    > + if (!u)
    > + /* No matching uprobe; signal SIGTRAP. */
    > + goto cleanup_ret;
    > + if (!utask) {
    > + utask = add_utask();
    > + /* Cannot Allocate; re-execute the instruction. */
    > + if (!utask)
    > + goto cleanup_ret;
    > + }
    > + /* TODO Start queueing signals. */
    > + utask->active_uprobe = u;
    > + handler_chain(u, regs);
    > + utask->state = UTASK_SSTEP;
    > + if (!pre_ssout(u, regs, probept))
    > + user_enable_single_step(current);

    Oooh. Playing with user_*_single_step() is obviously not very nice...
    But I guess you have no choice. Although I _hope_ we can do something
    else later.

    And what if we step into a syscall insn? I do not understand this
    low level code, but it seems that in this case we trap in kernel mode
    and do_debug() doesn't clear X86_EFLAGS_TF because uprobes hook
    DIE_DEBUG. IOW, the task will trap again and again inside this syscall,
    no?

    > + } else if (utask->state == UTASK_SSTEP) {
    > + u = utask->active_uprobe;
    > + if (sstep_complete(u, regs)) {

    It is not clear to me if it is correct to simply return if
    sstep_complete() returns false... What if X86_EFLAGS_TF was "lost"
    somehow?


    Again, I am not saying I understand this magic. Not at all ;)
    Please simply ignore my email if you think everything is fine.

    Oleg.



    \
     
     \ /
      Last update: 2011-10-07 20:35    [W:4.945 / U:0.320 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site