class FastbootRun:
def __init__(self, args, test_timeout):
self.powerup = args.powerup
self.ser = SerialBuffer(args.dev, "results/serial-output.txt",
"R SERIAL> ")
self.fastboot = "fastboot boot -s {ser} artifacts/fastboot.img".format(
ser=args.fbserial)
self.test_timeout = test_timeout
def close(self):
self.ser.close()
def print_error(self, message):
RED = '\033[0;31m'
NO_COLOR = '\033[0m'
print(RED + message + NO_COLOR)
def logged_system(self, cmd, timeout=60):
print("Running '{}'".format(cmd))
try:
return subprocess.call(cmd, shell=True, timeout=timeout)
except subprocess.TimeoutExpired:
self.print_error("timeout, restarting run...")
return 2
def run(self):
if ret := self.logged_system(self.powerup):
return ret
fastboot_ready = False
for line in self.ser.lines(timeout=2 * 60, phase="bootloader"):
if re.search("fastboot: processing commands", line) or \
re.search("Listening for fastboot command on", line):
fastboot_ready = True
break
if re.search("data abort", line):
self.print_error(
"Detected crash during boot, restarting run...")
return 2
if not fastboot_ready:
self.print_error(
"Failed to get to fastboot prompt, restarting run...")
return 2
if ret := self.logged_system(self.fastboot):
return ret
class CrosServoRun:
def __init__(self, cpu, ec):
# Merged FIFO for the two serial buffers, fed by threads.
self.serial_queue = queue.Queue()
self.sentinel = object()
self.threads_done = 0
self.ec_ser = SerialBuffer(ec, "results/serial-ec.txt",
"R SERIAL-EC> ")
self.cpu_ser = SerialBuffer(cpu, "results/serial.txt",
"R SERIAL-CPU> ")
self.iter_feed_ec = threading.Thread(target=self.iter_feed_queue,
daemon=True,
args=(self.ec_ser.lines(), ))
self.iter_feed_ec.start()
self.iter_feed_cpu = threading.Thread(target=self.iter_feed_queue,
daemon=True,
args=(self.cpu_ser.lines(), ))
self.iter_feed_cpu.start()
def close(self):
self.ec_ser.close()
self.cpu_ser.close()
self.iter_feed_ec.join()
self.iter_feed_cpu.join()
# Feed lines from our serial queues into the merged queue, marking when our
# input is done.
def iter_feed_queue(self, it):
for i in it:
self.serial_queue.put(i)
self.serial_queue.put(self.sentinel)
# Return the next line from the queue, counting how many threads have
# terminated and joining when done
def get_serial_queue_line(self):
line = self.serial_queue.get()
if line == self.sentinel:
self.threads_done = self.threads_done + 1
if self.threads_done == 2:
self.iter_feed_cpu.join()
self.iter_feed_ec.join()
return line
# Returns an iterator for getting the next line.
def serial_queue_lines(self):
return iter(self.get_serial_queue_line, self.sentinel)
def ec_write(self, s):
print("W SERIAL-EC> %s" % s)
self.ec_ser.serial.write(s.encode())
def cpu_write(self, s):
print("W SERIAL-CPU> %s" % s)
self.cpu_ser.serial.write(s.encode())
def print_error(self, message):
RED = '\033[0;31m'
NO_COLOR = '\033[0m'
print(RED + message + NO_COLOR)
def run(self):
# Flush any partial commands in the EC's prompt, then ask for a reboot.
self.ec_write("\n")
self.ec_write("reboot\n")
# This is emitted right when the bootloader pauses to check for input.
# Emit a ^N character to request network boot, because we don't have a
# direct-to-netboot firmware on cheza.
for line in self.serial_queue_lines():
if re.search("load_archive: loading locale_en.bin", line):
self.cpu_write("\016")
break
# The Cheza boards have issues with failing to bring up power to
# the system sometimes, possibly dependent on ambient temperature
# in the farm.
if re.search("POWER_GOOD not seen in time", line):
self.print_error(
"Detected intermittent poweron failure, restarting run...")
return 2
tftp_failures = 0
for line in self.serial_queue_lines():
if re.search("---. end Kernel panic", line):
return 1
# The Cheza firmware seems to occasionally get stuck looping in
# this error state during TFTP booting, possibly based on amount of
# network traffic around it, but it'll usually recover after a
# reboot.
if re.search("R8152: Bulk read error 0xffffffbf", line):
tftp_failures += 1
if tftp_failures >= 100:
self.print_error(
"Detected intermittent tftp failure, restarting run..."
)
return 2
# There are very infrequent bus errors during power management transitions
# on cheza, which we don't expect to be the case on future boards.
if re.search(
"Kernel panic - not syncing: Asynchronous SError Interrupt",
line):
self.print_error(
"Detected cheza power management bus error, restarting run..."
)
return 2
# If the network device dies, it's probably not graphics's fault, just try again.
if re.search("NETDEV WATCHDOG", line):
self.print_error(
"Detected network device failure, restarting run...")
return 2
# These HFI response errors started appearing with the introduction
# of piglit runs. CosmicPenguin says:
#
# "message ID 106 isn't a thing, so likely what happened is that we
# got confused when parsing the HFI queue. If it happened on only
# one run, then memory corruption could be a possible clue"
#
# Given that it seems to trigger randomly near a GPU fault and then
# break many tests after that, just restart the whole run.
if re.search(
"a6xx_hfi_send_msg.*Unexpected message id .* on the response queue",
line):
self.print_error(
"Detected cheza power management bus error, restarting run..."
)
return 2
if re.search("coreboot.*bootblock starting", line):
self.print_error(
"Detected spontaneous reboot, restarting run...")
return 2
if re.search(
"arm-smmu 5040000.iommu: TLB sync timed out -- SMMU may be deadlocked",
line):
self.print_error("Detected cheza MMU fail, restarting run...")
return 2
result = re.search("hwci: mesa: (\S*)", line)
if result:
if result.group(1) == "pass":
return 0
else:
return 1
self.print_error(
"Reached the end of the CPU serial log without finding a result")
return 1
class CrosServoRun:
def __init__(self, cpu, ec, test_timeout):
self.cpu_ser = SerialBuffer(cpu, "results/serial.txt",
"R SERIAL-CPU> ")
# Merge the EC serial into the cpu_ser's line stream so that we can
# effectively poll on both at the same time and not have to worry about
self.ec_ser = SerialBuffer(ec,
"results/serial-ec.txt",
"R SERIAL-EC> ",
line_queue=self.cpu_ser.line_queue)
self.test_timeout = test_timeout
def close(self):
self.ec_ser.close()
self.cpu_ser.close()
def ec_write(self, s):
print("W SERIAL-EC> %s" % s)
self.ec_ser.serial.write(s.encode())
def cpu_write(self, s):
print("W SERIAL-CPU> %s" % s)
self.cpu_ser.serial.write(s.encode())
def print_error(self, message):
RED = '\033[0;31m'
NO_COLOR = '\033[0m'
print(RED + message + NO_COLOR)
def run(self):
# Flush any partial commands in the EC's prompt, then ask for a reboot.
self.ec_write("\n")
self.ec_write("reboot\n")
bootloader_done = False
# This is emitted right when the bootloader pauses to check for input.
# Emit a ^N character to request network boot, because we don't have a
# direct-to-netboot firmware on cheza.
for line in self.cpu_ser.lines(timeout=120, phase="bootloader"):
if re.search("load_archive: loading locale_en.bin", line):
self.cpu_write("\016")
bootloader_done = True
break
# If the board has a netboot firmware and we made it to booting the
# kernel, proceed to processing of the test run.
if re.search("Booting Linux", line):
bootloader_done = True
break
# The Cheza boards have issues with failing to bring up power to
# the system sometimes, possibly dependent on ambient temperature
# in the farm.
if re.search("POWER_GOOD not seen in time", line):
self.print_error(
"Detected intermittent poweron failure, restarting run...")
return 2
if not bootloader_done:
print("Failed to make it through bootloader, restarting run...")
return 2
tftp_failures = 0
for line in self.cpu_ser.lines(timeout=self.test_timeout,
phase="test"):
if re.search("---. end Kernel panic", line):
return 1
# The Cheza firmware seems to occasionally get stuck looping in
# this error state during TFTP booting, possibly based on amount of
# network traffic around it, but it'll usually recover after a
# reboot.
if re.search("R8152: Bulk read error 0xffffffbf", line):
tftp_failures += 1
if tftp_failures >= 100:
self.print_error(
"Detected intermittent tftp failure, restarting run..."
)
return 2
# There are very infrequent bus errors during power management transitions
# on cheza, which we don't expect to be the case on future boards.
if re.search(
"Kernel panic - not syncing: Asynchronous SError Interrupt",
line):
self.print_error(
"Detected cheza power management bus error, restarting run..."
)
return 2
# If the network device dies, it's probably not graphics's fault, just try again.
if re.search("NETDEV WATCHDOG", line):
self.print_error(
"Detected network device failure, restarting run...")
return 2
# These HFI response errors started appearing with the introduction
# of piglit runs. CosmicPenguin says:
#
# "message ID 106 isn't a thing, so likely what happened is that we
# got confused when parsing the HFI queue. If it happened on only
# one run, then memory corruption could be a possible clue"
#
# Given that it seems to trigger randomly near a GPU fault and then
# break many tests after that, just restart the whole run.
if re.search(
"a6xx_hfi_send_msg.*Unexpected message id .* on the response queue",
line):
self.print_error(
"Detected cheza power management bus error, restarting run..."
)
return 2
if re.search("coreboot.*bootblock starting", line):
self.print_error(
"Detected spontaneous reboot, restarting run...")
return 2
if re.search(
"arm-smmu 5040000.iommu: TLB sync timed out -- SMMU may be deadlocked",
line):
self.print_error("Detected cheza MMU fail, restarting run...")
return 2
result = re.search("hwci: mesa: (\S*)", line)
if result:
if result.group(1) == "pass":
return 0
else:
return 1
self.print_error(
"Reached the end of the CPU serial log without finding a result")
return 2
class FastbootRun:
def __init__(self, args):
self.powerup = args.powerup
# We would like something like a 1 minute timeout, but the piglit traces
# jobs stall out for long periods of time.
self.ser = SerialBuffer(args.dev,
"results/serial-output.txt",
"R SERIAL> ",
timeout=600)
self.fastboot = "fastboot boot -s {ser} artifacts/fastboot.img".format(
ser=args.fbserial)
def close(self):
self.ser.close()
def print_error(self, message):
RED = '\033[0;31m'
NO_COLOR = '\033[0m'
print(RED + message + NO_COLOR)
def logged_system(self, cmd):
print("Running '{}'".format(cmd))
return os.system(cmd)
def run(self):
if self.logged_system(self.powerup) != 0:
return 1
fastboot_ready = False
for line in self.ser.lines():
if re.search("fastboot: processing commands", line) or \
re.search("Listening for fastboot command on", line):
fastboot_ready = True
break
if re.search("data abort", line):
self.print_error(
"Detected crash during boot, restarting run...")
return 2
if not fastboot_ready:
self.print_error(
"Failed to get to fastboot prompt, restarting run...")
return 2
if self.logged_system(self.fastboot) != 0:
return 1
print_more_lines = -1
for line in self.ser.lines():
if print_more_lines == 0:
return 2
if print_more_lines > 0:
print_more_lines -= 1
if re.search("---. end Kernel panic", line):
return 1
# The db820c boards intermittently reboot. Just restart the run
# when if we see a reboot after we got past fastboot.
if re.search("PON REASON", line):
self.print_error(
"Detected spontaneous reboot, restarting run...")
return 2
# db820c sometimes wedges around iommu fault recovery
if re.search("watchdog: BUG: soft lockup - CPU.* stuck", line):
self.print_error(
"Detected kernel soft lockup, restarting run...")
return 2
# If the network device dies, it's probably not graphics's fault, just try again.
if re.search("NETDEV WATCHDOG", line):
self.print_error(
"Detected network device failure, restarting run...")
return 2
# A3xx recovery doesn't quite work. Sometimes the GPU will get
# wedged and recovery will fail (because power can't be reset?)
# This assumes that the jobs are sufficiently well-tested that GPU
# hangs aren't always triggered, so just try again. But print some
# more lines first so that we get better information on the cause
# of the hang. Once a hang happens, it's pretty chatty.
if "[drm:adreno_recover] *ERROR* gpu hw init failed: -22" in line:
self.print_error("Detected GPU hang, restarting run...")
if print_more_lines == -1:
print_more_lines = 30
result = re.search("hwci: mesa: (\S*)", line)
if result:
if result.group(1) == "pass":
return 0
else:
return 1
self.print_error(
"Reached the end of the CPU serial log without finding a result, restarting run..."
)
return 2