Thread (119 messages) 119 messages, 10 authors, 2011-07-25

Re: [PATCH v4 3.0-rc2-tip 2/22] 2: uprobes: Breakground page replacement.

From: Peter Zijlstra <peterz@infradead.org>
Date: 2011-06-14 12:58:46
Also in: lkml

On Mon, 2011-06-13 at 14:29 +0530, Srikar Dronamraju wrote:
* Peter Zijlstra [off-list ref] [2011-06-10 01:03:29]:
quoted
On Tue, 2011-06-07 at 18:28 +0530, Srikar Dronamraju wrote:
quoted
+       vaddr_old = kmap_atomic(old_page, KM_USER0);
+       vaddr_new = kmap_atomic(new_page, KM_USER1);
+
+       memcpy(vaddr_new, vaddr_old, PAGE_SIZE);
+       /* poke the new insn in, ASSUMES we don't cross page boundary */
+       addr = vaddr;
+       vaddr &= ~PAGE_MASK;
+       memcpy(vaddr_new + vaddr, &opcode, uprobe_opcode_sz);
+
+       kunmap_atomic(vaddr_new);
+       kunmap_atomic(vaddr_old); 
quoted
+       vaddr_new = kmap_atomic(page, KM_USER0);
+       vaddr &= ~PAGE_MASK;
+       memcpy(opcode, vaddr_new + vaddr, uprobe_opcode_sz);
+       kunmap_atomic(vaddr_new);
quoted
Both sequences in resp {write,read}_opcode() assume the opcode doesn't
cross page boundaries but don't in fact have any assertions validating
this assumption.
read_opcode and write_opcode reads/writes just one breakpoint instruction
I had the below note just above the write_opcode definition.

/*
 * NOTE:
 * Expect the breakpoint instruction to be the smallest size instruction for
 * the architecture. If an arch has variable length instruction and the
 * breakpoint instruction is not of the smallest length instruction
 * supported by that architecture then we need to modify read_opcode /
 * write_opcode accordingly. This would never be a problem for archs that
 * have fixed length instructions.
 */
Whoever reads comments anyway? :-)
Do we have archs which have a breakpoint instruction which isnt of the
smallest instruction size for that arch. If we do have can we change the
write_opcode/read_opcode while we support that architecture?
Why not put a simple WARN_ON_ONCE() in there that checks the assumption?

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help