Re: [Qemu-devel] [PATCH 4/9] ATAPI: Replace DEBUG_IDE_ATAPI with tracing events
On 08/08/2017 04:59 PM, Eric Blake wrote:
> On 08/08/2017 01:33 PM, John Snow wrote:
>> Goodbye, printfs.
>> Hello, fancy printfs.
>>
>> Signed-off-by: John Snow
>> ---
>> hw/ide/atapi.c| 64
>> +--
>> hw/ide/trace-events | 19 ++
>> include/hw/ide/internal.h | 1 -
>> 3 files changed, 42 insertions(+), 42 deletions(-)
>
>> @@ -1330,16 +1310,18 @@ void ide_atapi_cmd(IDEState *s)
>> uint8_t *buf = s->io_buffer;
>> const struct AtapiCmd *cmd = &atapi_cmd_table[s->io_buffer[0]];
>>
>> -#ifdef DEBUG_IDE_ATAPI
>> -{
>> +trace_ide_atapi_cmd(s, s->io_buffer[0]);
>> +
>> +if (TRACE_IDE_ATAPI_CMD_PACKET_ENABLED) {
>> +/* Each pretty-printed byte needs two bytes and a space; */
>> +char *ppacket = g_malloc(ATAPI_PACKET_SIZE * 3 + 1);
>
> Nice use of an extra conditional to make a pretty trace without the
> overhead when tracing is off. (Oh yeah, you asked me on IRC how to do
> it, and I pointed to commit bd6952a3 as the example)
>
>
>> +++ b/hw/ide/trace-events
>> @@ -6,6 +6,10 @@ ide_ioport_read(uint32_t addr, const char *reg, uint32_t
>> val, void *bus, void *s
>> ide_ioport_write(uint32_t addr, const char *reg, uint32_t val, void *bus,
>> void *s) "IDE PIO wr @ 0x%"PRIx32" (%s); val 0x%02"PRIx32"; bus %p IDEState
>> %p"
>
> Wouldn't it be nice if our trace generator would let us line-wrap?
>
>> +# Warning: Verbose
>
> Cute. Should you also add it to the ide_data_* traces, per the commit
> message in 3/9?
>
Probably.
>> +ide_atapi_cmd_packet(void *s, uint16_t limit, const char *packet)
>> "IDEState: %p; limit=0x%x packet: %s"
>> \ No newline at end of file
>
> Umm, that should be fixed (I know that 'diff' uses both / and \ to
> differentiate whether it was the pre- or post-patch version that had the
> issue, but never remember which is which - so the fix may be in an
> earlier patch). Also fix the trace-events rebase for 3/9; with that,
> you can add
>
I'll fix these and let you re-bestow me with RBs, to err on the side of
caution.
> Reviewed-by: Eric Blake
>
Re: [Qemu-devel] [PATCH 4/9] ATAPI: Replace DEBUG_IDE_ATAPI with tracing events
On 08/08/2017 01:33 PM, John Snow wrote:
> Goodbye, printfs.
> Hello, fancy printfs.
>
> Signed-off-by: John Snow
> ---
> hw/ide/atapi.c| 64
> +--
> hw/ide/trace-events | 19 ++
> include/hw/ide/internal.h | 1 -
> 3 files changed, 42 insertions(+), 42 deletions(-)
> @@ -1330,16 +1310,18 @@ void ide_atapi_cmd(IDEState *s)
> uint8_t *buf = s->io_buffer;
> const struct AtapiCmd *cmd = &atapi_cmd_table[s->io_buffer[0]];
>
> -#ifdef DEBUG_IDE_ATAPI
> -{
> +trace_ide_atapi_cmd(s, s->io_buffer[0]);
> +
> +if (TRACE_IDE_ATAPI_CMD_PACKET_ENABLED) {
> +/* Each pretty-printed byte needs two bytes and a space; */
> +char *ppacket = g_malloc(ATAPI_PACKET_SIZE * 3 + 1);
Nice use of an extra conditional to make a pretty trace without the
overhead when tracing is off. (Oh yeah, you asked me on IRC how to do
it, and I pointed to commit bd6952a3 as the example)
> +++ b/hw/ide/trace-events
> @@ -6,6 +6,10 @@ ide_ioport_read(uint32_t addr, const char *reg, uint32_t
> val, void *bus, void *s
> ide_ioport_write(uint32_t addr, const char *reg, uint32_t val, void *bus,
> void *s) "IDE PIO wr @ 0x%"PRIx32" (%s); val 0x%02"PRIx32"; bus %p IDEState
> %p"
Wouldn't it be nice if our trace generator would let us line-wrap?
> +# Warning: Verbose
Cute. Should you also add it to the ide_data_* traces, per the commit
message in 3/9?
> +ide_atapi_cmd_packet(void *s, uint16_t limit, const char *packet) "IDEState:
> %p; limit=0x%x packet: %s"
> \ No newline at end of file
Umm, that should be fixed (I know that 'diff' uses both / and \ to
differentiate whether it was the pre- or post-patch version that had the
issue, but never remember which is which - so the fix may be in an
earlier patch). Also fix the trace-events rebase for 3/9; with that,
you can add
Reviewed-by: Eric Blake
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
signature.asc
Description: OpenPGP digital signature
