On Mon, 7 May 2012, Peter Korsgaard wrote: > Hi, > > There seems to be a mismatch with the behaviour of dummy_hcd and > composite regarding removal after the move to new style udc_start/stop. > I get an oops with: > > modprobe dummy_hcd > modprobe g_serial > rmmod g_serial > > Process rmmod (pid: 4233, threadinfo ffff88006c7e6000, task ffff88006c78e930) > Stack: > ffffffffa00a026b ffff88006c73a9a0 ffffffffa004f8a0 ffffffffa0050300 > ffffffffa004c5de ffff880079fc43c0 ffffffffa0050330 ffff88006c73a9a0 > ffffffffa004f8a0 0000000000000000 ffffffffa004cade ffff88006c739400 > Call Trace: > [<ffffffffa00a026b>] ? composite_disconnect+0x3b/0x80 [g_serial] > [<ffffffffa004c5de>] ? set_link_state+0x20e/0x2f0 [dummy_hcd] > [<ffffffffa004cade>] ? dummy_pullup+0x10e/0x140 [dummy_hcd] > [<ffffffffa000304b>] ? usb_gadget_remove_driver+0x4b/0x80 [udc_core] > [<ffffffffa00030e4>] ? usb_gadget_unregister_driver+0x64/0x90 [udc_core] > [<ffffffffa00a364e>] ? cleanup+0xd/0x13 [g_serial] > [<ffffffffa00a3641>] ? override_id.isra.24+0x3d/0x3d [g_serial] > [<ffffffff8106849d>] ? sys_delete_module+0x16d/0x230 > [<ffffffff814c3522>] ? system_call_fastpath+0x16/0x1b > Code: 38 c2 74 10 0f 1f 80 00 00 00 00 f3 90 0f b6 07 38 d0 75 f7 c3 66 66 66 66 66 2e 0f 1f 84 00 00 00 00 00 9c 58 fa ba 00 01 00 00 <f0> 66 0f c1 17 0f b6 ce 38 d1 74 0d 0f 1f 40 00 f3 90 0f b6 17 > RIP [<ffffffff814c2ac8>] _raw_spin_lock_irqsave+0x8/0x30 > > Which is caused by usb_gadget_remove_driver calling composite_unbind(), > which clears the gadget data, and then usb_gadget_disconnect -> > dummy_pullup -> set_link_state which calls composite_disconnect which > then oopses as the gadget_data is gone. > > So who is to blame? Should we protect composite_disconnect with a cdev > test to adjust dummy_hcd to not call it? > > This is with 3.3.4, but I don't right away see any relevant changes in > git. This has been fixed in 3.4-rc5 by commit 320cd1e750f1bf3e47eb41209dcb2be07264cb76 (usb: gadget: udc-core: fix incompatibility with dummy-hcd). Are you suggesting that this patch (along with the two preceding changes to udc-core.c) should be pack-ported to the stable kernels? Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
