Re: [Qemu-devel] [PATCH] qtest: add a fuzz test to fdc-test
Il 21/05/2012 09:51, Kevin Wolf ha scritto: GTESTER check-qtest-i386 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x1f FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xa8 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x37 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x93 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xe4 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xc1 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x92 Maybe they should be fixed first. What do you mean by fixing? Turning them into DPRINTFs? Or trace events? Paolo
Re: [Qemu-devel] [PATCH] qtest: add a fuzz test to fdc-test
Am 21.05.2012 10:11, schrieb Paolo Bonzini: Il 21/05/2012 09:51, Kevin Wolf ha scritto: GTESTER check-qtest-i386 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x1f FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xa8 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x37 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x93 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xe4 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xc1 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x92 Maybe they should be fixed first. What do you mean by fixing? Turning them into DPRINTFs? Or trace events? Yeah, you could turn all FLOPPY_DPRINTFs into trace events. But the point here is that today it's a FLOPPY_ERROR, and except for register fuzzing they report real problems with the emulation and not just some debugging information. So I'm not sure if hiding them is really a fix. Kevin
Re: [Qemu-devel] [PATCH] qtest: add a fuzz test to fdc-test
What do you mean by fixing? Turning them into DPRINTFs? Or trace events? Yeah, you could turn all FLOPPY_DPRINTFs into trace events. But the point here is that today it's a FLOPPY_ERROR, and except for register fuzzing they report real problems with the emulation and not just some debugging information. So I'm not sure if hiding them is really a fix. It depends, controller not ready for reading is most likely just caused by fuzzing. Most unimplemented commands are also invalid on real hardware too. Paolo
Re: [Qemu-devel] [PATCH] qtest: add a fuzz test to fdc-test
Am 19.05.2012 14:54, schrieb Blue Swirl: Add a simple register fuzzing test to floppy controller tests. Signed-off-by: Blue Swirl blauwir...@gmail.com --- There's a lot of output like: GTESTER check-qtest-i386 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x1f FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xa8 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x37 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x93 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xe4 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xc1 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x92 Maybe they should be fixed first. What do you mean by fixing? Turning them into DPRINTFs? Kevin
Re: [Qemu-devel] [PATCH] qtest: add a fuzz test to fdc-test
On Mon, May 21, 2012 at 8:14 AM, Kevin Wolf kw...@redhat.com wrote: Am 21.05.2012 10:11, schrieb Paolo Bonzini: Il 21/05/2012 09:51, Kevin Wolf ha scritto: GTESTER check-qtest-i386 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x1f FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xa8 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x37 FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_read_data: controller not ready for reading FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x93 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xe4 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0xc1 FLOPPY ERROR: fdctrl_unimplemented: unimplemented command 0x92 Maybe they should be fixed first. What do you mean by fixing? Turning them into DPRINTFs? Or trace events? Yeah, you could turn all FLOPPY_DPRINTFs into trace events. But the point here is that today it's a FLOPPY_ERROR, and except for register fuzzing they report real problems with the emulation and not just some debugging information. So I'm not sure if hiding them is really a fix. While not a DoS, letting the guest spam the console at will is not nice either. Maybe we need a new method to enable a selected set of printouts, something like '-d unimplemented'. That way no recompiling would be needed. Kevin
Re: [Qemu-devel] [PATCH] qtest: add a fuzz test to fdc-test
On Mon, May 21, 2012 at 8:18 AM, Paolo Bonzini pbonz...@redhat.com wrote: What do you mean by fixing? Turning them into DPRINTFs? Or trace events? Yeah, you could turn all FLOPPY_DPRINTFs into trace events. But the point here is that today it's a FLOPPY_ERROR, and except for register fuzzing they report real problems with the emulation and not just some debugging information. So I'm not sure if hiding them is really a fix. It depends, controller not ready for reading is most likely just caused by fuzzing. Most unimplemented commands are also invalid on real hardware too. Yes, but a malevolent guest could issue the same commands if those could cause problems to QEMU. Paolo
Re: [Qemu-devel] [PATCH] qtest: add a fuzz test to fdc-test
On 21 May 2012 18:30, Blue Swirl blauwir...@gmail.com wrote: On Mon, May 21, 2012 at 8:14 AM, Kevin Wolf kw...@redhat.com wrote: Yeah, you could turn all FLOPPY_DPRINTFs into trace events. But the point here is that today it's a FLOPPY_ERROR, and except for register fuzzing they report real problems with the emulation and not just some debugging information. So I'm not sure if hiding them is really a fix. While not a DoS, letting the guest spam the console at will is not nice either. Maybe we need a new method to enable a selected set of printouts, something like '-d unimplemented'. That way no recompiling would be needed. +1 for a better set of graduated logging/debug levels and a sensible command line interface for turning them on and off. Possible severity levels: * debug output * guest has done something that suggests it might be buggy, eg accessing nonexistent register * guest has tried to use something qemu doesn't implement * qemu (fatal) error These should be orthogonal to the what area should we print logging for question, I think. With a clean API for defining log messages I think we could clean up a lot of the legacy functions for asserting/aborting in various ways (in particular a lot of the hw_error() uses). -- PMM
