[device/Arista] Improvements to the boot of Arista devices. (#2898)

* Fix showing systemd shutdown sequence when verbose is set

* Fix creation of kernel-cmdline file

Sometimes boot0 prints error
"mv: can't preserve ownership of '/mnt/flash/image-arsonic.xxxx/kernel-cmdline': Operation not permitted"

* Improve flash space usage during installation

Some older systems only have 2GB of flash available. Installing a second
image on these can prove to be challenging.
The new installation process moves the installer swi to memory in order
to avoid free up space from the flash before uncompressing it there.
It removes all the flash space usage spike and also improves the IO
since the installation is no more reading and writting to the flash at
the same time.

* Add support of 7060CX-32S-SSD

* 7260CX3: use inventory powerCycle procedures

* 7050QX-32S: use inventory powerCycle procedures

* 7050QX-32: use inventory powerCycle procedures

* platform: arista: add common platform_reboot

Replace platform_reboot by a link to new common for devices already
using a similar script.

* 7060CX-32S: use inventory powerCycle procedures

* Install python smbus in pmon

Some platform plugin need the python smbus library to perform some actions.
This installs the dependency.

device/arista/x86_64-arista_7050_qx32/platform_reboot

@@ -1,87 +0,0 @@
-#!/usr/bin/env python
-# Copyright (c) 2018 Arista Networks, Inc.  All rights reserved.
-# Arista Networks, Inc. Confidential and Proprietary.
-
-# Reboot script for 7050QX-32
-
-from __future__ import print_function
-import sys
-import mmap, os
-import subprocess
-from struct import pack, unpack
-
-class MmapResource( object ):
-   """Resource implementation for a directly-mapped memory region."""
-
-   def __init__( self, path ):
-      try:
-         fd = os.open( path, os.O_RDWR )
-      except EnvironmentError:
-         print( "FAIL can not open scd memory-map resource file" )
-         print( "FAIL are you running on the proper platform?" )
-         sys.exit( 1 )
-      try:
-         size = os.fstat( fd ).st_size
-      except EnvironmentError:
-         print( "FAIL can not fstat scd memory-map resource file" )
-         print( "FAIL are you running on the proper platform?" )
-         sys.exit( 1 )
-      try:
-         self.mmap_ = mmap.mmap( fd, size, mmap.MAP_SHARED,
-                      mmap.PROT_READ | mmap.PROT_WRITE )
-      except EnvironmentError:
-         print( "FAIL can not map scd memory-map file" )
-         print( "FAIL are you running on the proper platform?" )
-         sys.exit( 1 )
-      finally:
-         try:
-            # Note that closing the file descriptor has no effect on the memory map
-            os.close( fd )
-         except EnvironmentError:
-            print( "FAIL failed to close scd memory-map file" )
-            sys.exit( 1 )
-
-   def read32( self, addr ):
-      return unpack( '<L', self.mmap_[ addr : addr + 4 ] )[ 0 ]
-
-   def write32( self, addr, value ):
-      self.mmap_[ addr: addr + 4 ] = pack( '<L', value )
-
-def scdRegTest( scd, offset, val1, count ):
-   scd.write32( offset, val1 )
-   val2 = scd.read32( offset )
-   if val1 != val2:
-      print( "FAIL: scd write 0x%08x but read back 0x%08x in iter %d" %
-             ( val1, val2, count ) )
-      sys.exit( 17 )
-
-def scdScrRegTest( scd ):
-   scrOffset = 0x0130
-   for i in range( 0, 3 ):
-      scdRegTest( scd, scrOffset, 0xdeadbeef, i )
-      scdRegTest( scd, scrOffset, 0xa5a5a5a5, i )
-      scdRegTest( scd, scrOffset, 0x00000000, i )
-
-def reboot( scd ):
-   # reboot the system by writing to register 0x7000
-   print( "Rebooting" )
-   scd.write32( 0x7000, 0xDEAD )
-   print( "REBOOTED" )
-
-def main():
-   busName = "/sys/bus/pci/devices/0000:04:00.0/resource0"
-
-   subprocess.call( [ 'modprobe', 'scd' ] )
-
-   scd = MmapResource( busName )
-
-   #
-   # verify that we can read/write scd scratch register
-   #
-   scdScrRegTest( scd )
-
-   # reboot the system
-   reboot( scd )
-
-if __name__ == "__main__":
-   main()

device/arista/x86_64-arista_7050_qx32/platform_reboot

@@ -0,0 +1 @@
+../x86_64-arista_common/platform_reboot

device/arista/x86_64-arista_7050_qx32s/platform_reboot

@@ -1,87 +0,0 @@
-#!/usr/bin/env python
-# Copyright (c) 2018 Arista Networks, Inc.  All rights reserved.
-# Arista Networks, Inc. Confidential and Proprietary.
-
-# Reboot script for 7050QX-32S
-
-from __future__ import print_function
-import sys
-import mmap, os
-import subprocess
-from struct import pack, unpack
-
-class MmapResource( object ):
-   """Resource implementation for a directly-mapped memory region."""
-
-   def __init__( self, path ):
-      try:
-         fd = os.open( path, os.O_RDWR )
-      except EnvironmentError:
-         print( "FAIL can not open scd memory-map resource file" )
-         print( "FAIL are you running on the proper platform?" )
-         sys.exit( 1 )
-      try:
-         size = os.fstat( fd ).st_size
-      except EnvironmentError:
-         print( "FAIL can not fstat scd memory-map resource file" )
-         print( "FAIL are you running on the proper platform?" )
-         sys.exit( 1 )
-      try:
-         self.mmap_ = mmap.mmap( fd, size, mmap.MAP_SHARED,
-                      mmap.PROT_READ | mmap.PROT_WRITE )
-      except EnvironmentError:
-         print( "FAIL can not map scd memory-map file" )
-         print( "FAIL are you running on the proper platform?" )
-         sys.exit( 1 )
-      finally:
-         try:
-            # Note that closing the file descriptor has no effect on the memory map
-            os.close( fd )
-         except EnvironmentError:
-            print( "FAIL failed to close scd memory-map file" )
-            sys.exit( 1 )
-
-   def read32( self, addr ):
-      return unpack( '<L', self.mmap_[ addr : addr + 4 ] )[ 0 ]
-
-   def write32( self, addr, value ):
-      self.mmap_[ addr: addr + 4 ] = pack( '<L', value )
-
-def scdRegTest( scd, offset, val1, count ):
-   scd.write32( offset, val1 )
-   val2 = scd.read32( offset )
-   if val1 != val2:
-      print( "FAIL: scd write 0x%08x but read back 0x%08x in iter %d" %
-             ( val1, val2, count ) )
-      sys.exit( 17 )
-
-def scdScrRegTest( scd ):
-   scrOffset = 0x0130
-   for i in range( 0, 3 ):
-      scdRegTest( scd, scrOffset, 0xdeadbeef, i )
-      scdRegTest( scd, scrOffset, 0xa5a5a5a5, i )
-      scdRegTest( scd, scrOffset, 0x00000000, i )
-
-def reboot( scd ):
-   # reboot the system by writing to register 0x7000
-   print( "Rebooting" )
-   scd.write32( 0x7000, 0xDEAD )
-   print( "REBOOTED" )
-
-def main():
-   busName = "/sys/bus/pci/devices/0000:02:00.0/resource0"
-
-   subprocess.call( [ 'modprobe', 'scd' ] )
-
-   scd = MmapResource( busName )
-
-   #
-   # verify that we can read/write scd scratch register
-   #
-   scdScrRegTest( scd )
-
-   # reboot the system
-   reboot( scd )
-
-if __name__ == "__main__":
-   main()

device/arista/x86_64-arista_7050_qx32s/platform_reboot

@@ -0,0 +1 @@
+../x86_64-arista_common/platform_reboot

device/arista/x86_64-arista_7060_cx32s/platform_reboot

@@ -1,18 +0,0 @@
-#!/usr/bin/env python
-# Copyright (c) 2018 Arista Networks, Inc.  All rights reserved.
-# Arista Networks, Inc. Confidential and Proprietary.
-
-# Reboot script for 7060CX-32
-
-from __future__ import print_function
-
-import smbus
-
-def main():
-   print( "Rebooting" )
-   bus = smbus.SMBus( 1 )
-   bus.write_byte_data( 0x23, 0x04, 0xde )
-   print( "REBOOTED" )
-
-if __name__ == "__main__":
-   main()

device/arista/x86_64-arista_7060_cx32s/platform_reboot

@@ -0,0 +1 @@
+../x86_64-arista_common/platform_reboot

device/arista/x86_64-arista_7060px4_32/platform_reboot

@@ -1,11 +0,0 @@
-#!/usr/bin/env python
-
-from arista import platforms
-from arista.utils.sonic_reboot import reboot
-
-def main():
-   # reboot the system
-   reboot()
-
-if __name__ == "__main__":
-   main()

device/arista/x86_64-arista_7060px4_32/platform_reboot

@@ -0,0 +1 @@
+../x86_64-arista_common/platform_reboot

device/arista/x86_64-arista_7170_64c/platform_reboot

@@ -1,11 +0,0 @@
-#!/usr/bin/env python
-
-import arista.platforms
-from arista.utils.sonic_reboot import reboot
-
-def main():
-   # reboot the system
-   reboot()
-
-if __name__ == "__main__":
-   main()

device/arista/x86_64-arista_7170_64c/platform_reboot

@@ -0,0 +1 @@
+../x86_64-arista_common/platform_reboot

device/arista/x86_64-arista_7260cx3_64/platform_reboot

@@ -1,87 +0,0 @@
-#!/usr/bin/env python
-# Copyright (c) 2018 Arista Networks, Inc.  All rights reserved.
-# Arista Networks, Inc. Confidential and Proprietary.
-
-# Reboot script for 7260CX3
-
-from __future__ import print_function
-import sys
-import mmap, os
-import subprocess
-from struct import pack, unpack
-
-class MmapResource( object ):
-   """Resource implementation for a directly-mapped memory region."""
-
-   def __init__( self, path ):
-      try:
-         fd = os.open( path, os.O_RDWR )
-      except EnvironmentError:
-         print( "FAIL can not open scd memory-map resource file" )
-         print( "FAIL are you running on the proper platform?" )
-         sys.exit( 1 )
-      try:
-         size = os.fstat( fd ).st_size
-      except EnvironmentError:
-         print( "FAIL can not fstat scd memory-map resource file" )
-         print( "FAIL are you running on the proper platform?" )
-         sys.exit( 1 )
-      try:
-         self.mmap_ = mmap.mmap( fd, size, mmap.MAP_SHARED,
-                      mmap.PROT_READ | mmap.PROT_WRITE )
-      except EnvironmentError:
-         print( "FAIL can not map scd memory-map file" )
-         print( "FAIL are you running on the proper platform?" )
-         sys.exit( 1 )
-      finally:
-         try:
-            # Note that closing the file descriptor has no effect on the memory map
-            os.close( fd )
-         except EnvironmentError:
-            print( "FAIL failed to close scd memory-map file" )
-            sys.exit( 1 )
-
-   def read32( self, addr ):
-      return unpack( '<L', self.mmap_[ addr : addr + 4 ] )[ 0 ]
-
-   def write32( self, addr, value ):
-      self.mmap_[ addr: addr + 4 ] = pack( '<L', value )
-
-def scdRegTest( scd, offset, val1, count ):
-   scd.write32( offset, val1 )
-   val2 = scd.read32( offset )
-   if val1 != val2:
-      print( "FAIL: scd write 0x%08x but read back 0x%08x in iter %d" %
-             ( val1, val2, count ) )
-      sys.exit( 17 )
-
-def scdScrRegTest( scd ):
-   scrOffset = 0x0130
-   for i in range( 0, 3 ):
-      scdRegTest( scd, scrOffset, 0xdeadbeef, i )
-      scdRegTest( scd, scrOffset, 0xa5a5a5a5, i )
-      scdRegTest( scd, scrOffset, 0x00000000, i )
-
-def reboot( scd ):
-   # reboot the system by writing to register 0x7000
-   print( "Rebooting" )
-   scd.write32( 0x7000, 0xDEAD )
-   print( "REBOOTED" )
-
-def main():
-   busName = "/sys/bus/pci/devices/0000:ff:0b.3/resource0"
-
-   subprocess.call( [ 'modprobe', 'scd' ] )
-
-   scd = MmapResource( busName )
-
-   #
-   # verify that we can read/write scd scratch register
-   #
-   scdScrRegTest( scd )
-
-   # reboot the system
-   reboot( scd )
-
-if __name__ == "__main__":
-   main()

device/arista/x86_64-arista_7260cx3_64/platform_reboot

@@ -0,0 +1 @@
+../x86_64-arista_common/platform_reboot

device/arista/x86_64-arista_common/platform_reboot

@@ -0,0 +1,11 @@
+#!/usr/bin/env python
+
+import arista.platforms
+from arista.utils.sonic_reboot import reboot
+
+def main():
+   # reboot the system
+   reboot()
+
+if __name__ == "__main__":
+   main()

dockers/docker-platform-monitor/Dockerfile.j2

@@ -7,7 +7,7 @@ RUN [ -f /etc/rsyslog.conf ] && sed -ri "s/%syslogtag%/$docker_container_name#%s
 ENV DEBIAN_FRONTEND=noninteractive
 
 # Install required packages
-RUN apt-get update && apt-get install -y python-pip libpython2.7 ipmitool librrd8 librrd-dev rrdtool
+RUN apt-get update && apt-get install -y python-pip libpython2.7 ipmitool librrd8 librrd-dev rrdtool python-smbus
 
 {% if docker_platform_monitor_debs.strip() -%}
 # Copy locally-built Debian package dependencies

files/Aboot/boot0.j2

@@ -68,6 +68,8 @@ cmdline_base="$target_path/kernel-params-base"
 cmdline_image="$image_path/kernel-cmdline"
 boot_config="$target_path/boot-config"
 
+swi_tmpfs="/tmp/tmp-swi"
+
 bootconfigvars="KERNEL INITRD CONSOLESPEED PASSWORD NETDEV NETAUTO NETIP NETMASK NETGW NETDOMAIN NETDNS NETHW memtest"
 flash_re=" /mnt/flash| /host"
 
@@ -126,9 +128,34 @@ update_next_boot() {
     fi
 }
 
+move_swi_to_tmpfs() {
+    local oldswi="$1"
+    local newswi="$swi_tmpfs/$(basename $oldswi)"
+
+    mkdir -p "$swi_tmpfs"
+    if ! $in_aboot && ! mount | grep -q ' /tmp type tmpfs'; then
+        # mount a real tmpfs on /tmp/tmp-swi if /tmp is not one already.
+        mount -t tmpfs tmp-swi "$swi_tmpfs"
+    fi
+
+    mv "$oldswi" "$newswi"
+    echo "$newswi"
+}
+
+cleanup_swi_tmpfs() {
+   rm -f "$swipath"
+   if mount | grep -q "$swi_tmpfs"; then
+      umount "$swi_tmpfs" || :
+   fi
+}
+
 extract_image() {
     mkdir -p "$image_path"
 
+    info "Moving swi to a tmpfs"
+    ## Avoid problematic flash usage spike on older systems, also improves I/O
+    swipath="$(move_swi_to_tmpfs $swipath)"
+
     info "Extracting swi content"
     ## Unzip the image except boot0 and dockerfs archive
     unzip -oq "$swipath" -x boot0 "$dockerfs" -d "$image_path"
@@ -146,7 +173,6 @@ extract_image() {
         fi
 
         ## extract docker archive
-        info "Unpacking $dockerfs"
         unzip -oqp "$swipath" "$dockerfs" | tar xzf - -C "$image_path/{{ DOCKERFS_DIR }}" $TAR_EXTRA_OPTION
     else
         ## save dockerfs archive in the image directory
@@ -154,14 +180,15 @@ extract_image() {
         info "Unpacking $dockerfs delayed to initrd because $target_path is $rootfs_type"
     fi
 
+    ## remove installer since it's not needed anymore
+    info "Remove installer"
+    cleanup_swi_tmpfs
+
     ## use new reduced-size boot swi
     local swi_boot_path="flash:$image_name/{{ ABOOT_BOOT_IMAGE }}"
     update_boot_config SWI "$swi_boot_path"
     update_boot_config SWI_DEFAULT "$swi_boot_path"
 
-    ## Remove installer swi as it has lots of redundunt contents
-    rm -f "$swipath"
-
     ## sync disk operations
     sync
 }
@@ -205,7 +232,8 @@ platform_specific() {
         flash_size=3700
         echo "modprobe.blacklist=radeon,sp5100_tco" >>/tmp/append
     fi
-    if [ "$sid" = "Upperlake" ] || [ "$sid" = "UpperlakeES" ]; then
+    if [ "$sid" = "Upperlake" ] || [ "$sid" = "UpperlakeES" ] ||
+       [ "$sid" = "UpperlakeSsd" ]; then
         aboot_machine=arista_7060_cx32s
         flash_size=3700
         echo "amd_iommu=off" >> /tmp/append
@@ -304,7 +332,7 @@ write_boot_configs() {
         fi
     fi
 
-    mv /tmp/append $cmdline_image
+    cat /tmp/append > $cmdline_image
     [ -e ${target_path}/machine.conf ] || write_machine_config
 }
 
@@ -313,7 +341,10 @@ run_kexec() {
     local kernel="${KERNEL:-$(find $image_path/boot -name 'vmlinuz-*' -type f | head -n 1)}"
     local initrd="${INITRD:-$(find $image_path/boot -name 'initrd.img-*' -type f | head -n 1)}"
 
-    if ! $verbose; then
+    if $verbose; then
+       # show systemd showdown sequence when verbose is set
+       cmdline="$cmdline systemd.show_status=true"
+    else
        # Start showing systemd information from the first failing unit if any.
        # systemd.show_status=false or quiet can be used to silence systemd entierly
        cmdline="$cmdline systemd.show_status=auto"