RE: [PATCH 09/14] powerpc/vas: Update CSB and notify process for fault CRBs
flat view
From: Haren Myneni <hidden>
Date: 2019-12-06 22:49:04
<div class="socmaildefaultfont" dir="ltr" style="font-family:Arial, Helvetica, sans-serif;font-size:10pt" ><div dir="ltr" ><font face="Default Monospace,Courier New,Courier,monospace" size="2" >> <br>> +static void notify_process(pid_t pid, u64 fault_addr)<br>> +{<br>> + int rc;<br>> + struct kernel_siginfo info;<br>> +<br>> + memset(&info, 0, sizeof(info));<br>> +<br>> + info.si_signo = SIGSEGV;<br>> + info.si_errno = EFAULT;<br>> + info.si_code = SEGV_MAPERR;<br>> +<br>> + info.si_addr = (void *)fault_addr;<br>> + rcu_read_lock();<br>> + rc = kill_pid_info(SIGSEGV, &info, find_vpid(pid));<br>> + rcu_read_unlock();<br>> +<br>> + pr_devel("%s(): pid %d kill_proc_info() rc %d\n", __func__, pid, rc);<br>> +}<br><br>Shouldn't this use force_sig_fault_to_task instead?</font><br> </div>
<div dir="ltr" ><font face="Default Monospace,Courier New,Courier,monospace" size="2" >User space send compression request as Coprocessor Request Block (CRB) with CSB (status block)to NX directly and polls on CSB address. NX updates CSB after processing the request. If NX sees fault on the request buffer or CSB, interrupts kernel. So generally CSB should be valid since it supposed to poll csb.flags. </font></div>
<div dir="ltr" > </div>
<div dir="ltr" ><font face="Default Monospace,Courier New,Courier,monospace" size="2" >If an application does not behave as expected or NX pastes wrong information, an error message will be printed and send signal to an application. So applications or libnxz can ignore SEGV signal or can have their own handlers. </font><br> </div></div><BR>