|
| 1 | +- commits: |
| 2 | + - subject: Update layers/meta-balena to e6c2037615db74af70250c5d3351c44e2ee31fbb |
| 3 | + hash: ec4bf2e6466db15f888a906b92b88ac5a524e204 |
| 4 | + body: Update layers/meta-balena |
| 5 | + footer: |
| 6 | + Changelog-entry: Update layers/meta-balena to e6c2037615db74af70250c5d3351c44e2ee31fbb |
| 7 | + changelog-entry: Update layers/meta-balena to e6c2037615db74af70250c5d3351c44e2ee31fbb |
| 8 | + author: Self-hosted Renovate Bot |
| 9 | + nested: |
| 10 | + - commits: |
| 11 | + - subject: Update balena-supervisor to v16.3.17 |
| 12 | + hash: 521d51db1023846cf95e40a9f2b0a514ec8aea6d |
| 13 | + body: | |
| 14 | + Update balena-supervisor from 16.3.15 to 16.3.17 |
| 15 | + footer: |
| 16 | + Change-type: patch |
| 17 | + change-type: patch |
| 18 | + author: Self-hosted Renovate Bot |
| 19 | + nested: |
| 20 | + - commits: |
| 21 | + - subject: Fix engine deadlock on network+service change |
| 22 | + hash: ede27b63cecc2a9ae3139a905d9c6ab32418a72b |
| 23 | + body: > |
| 24 | + This fixes a regression on the supervisor state engine |
| 25 | + computation |
| 26 | + |
| 27 | + (added on v16.2.0) when |
| 28 | + |
| 29 | + the target state removes a network at the same time that |
| 30 | + a service |
| 31 | + |
| 32 | + referencing that network is changed. Example going from |
| 33 | + |
| 34 | + |
| 35 | + ``` |
| 36 | + |
| 37 | + services: |
| 38 | + one: |
| 39 | + image: alpine: 3.18 |
| 40 | + networks: ['balena'] |
| 41 | + |
| 42 | + networks: |
| 43 | + balena: |
| 44 | + ``` |
| 45 | + |
| 46 | + |
| 47 | + to |
| 48 | + |
| 49 | + |
| 50 | + ``` |
| 51 | + |
| 52 | + services: |
| 53 | + one: |
| 54 | + image: alpine: latest |
| 55 | + ``` |
| 56 | + |
| 57 | + |
| 58 | + Would never reach the target state as killing the |
| 59 | + service in order to |
| 60 | + |
| 61 | + remove the network is prioritized, but one of the |
| 62 | + invariants in the target state calculation is |
| 63 | + |
| 64 | + to not kill any services until all images have been |
| 65 | + downloaded. These |
| 66 | + |
| 67 | + two instructions were in contradiction leading to a |
| 68 | + deadlock. |
| 69 | + |
| 70 | + |
| 71 | + The fix involves only adding removal steps for services |
| 72 | + depending on a |
| 73 | + |
| 74 | + changing network or volume if the service container is |
| 75 | + not being removed |
| 76 | + |
| 77 | + already. |
| 78 | + footer: |
| 79 | + Change-type: patch |
| 80 | + change-type: patch |
| 81 | + author: Felipe Lalanne |
| 82 | + nested: [] |
| 83 | + version: balena-supervisor-16.3.17 |
| 84 | + title: "" |
| 85 | + date: 2024-06-25T01:03:26.734Z |
| 86 | + - commits: |
| 87 | + - subject: Update balena-io/deploy-to-balena-action action to v2.0.72 |
| 88 | + hash: 170733422a7a8fb2e58482508fe74dc2fcb282a5 |
| 89 | + body: > |
| 90 | + Update balena-io/deploy-to-balena-action from 2.0.71 to |
| 91 | + 2.0.72 |
| 92 | + footer: |
| 93 | + Change-type: patch |
| 94 | + change-type: patch |
| 95 | + author: Self-hosted Renovate Bot |
| 96 | + nested: [] |
| 97 | + version: balena-supervisor-16.3.16 |
| 98 | + title: "" |
| 99 | + date: 2024-06-17T01:56:08.257Z |
| 100 | + version: meta-balena-5.3.23 |
| 101 | + title: "" |
| 102 | + date: 2024-07-02T14:03:58.762Z |
| 103 | + - commits: |
| 104 | + - subject: "initrdscripts: make the kexec script fail hard in unexpected states" |
| 105 | + hash: 0b6199dddbb52e963d7b6e5feb5771f2b61a0f35 |
| 106 | + body: > |
| 107 | + At this moment the kexec initrd script is skipped when |
| 108 | + ROOTFS_DIR |
| 109 | + |
| 110 | + is not defined or if the new rootfs is mounted, but does not |
| 111 | + contain |
| 112 | + |
| 113 | + a kernel image in the expected place. This is undesirable as we |
| 114 | + |
| 115 | + assume this is the last script executed by the balena |
| 116 | + bootloader. |
| 117 | + |
| 118 | + |
| 119 | + This patch makes the kexec script always execute in the balena |
| 120 | + |
| 121 | + bootloader and makes it fail hard in unexpected states, which |
| 122 | + means |
| 123 | + |
| 124 | + the script is always an exit point for the balena bootloader, |
| 125 | + |
| 126 | + whether the actual kexec call succeeds or not. |
| 127 | + footer: |
| 128 | + Change-type: patch |
| 129 | + change-type: patch |
| 130 | + Signed-off-by: Michal Toman <michalt@balena.io> |
| 131 | + signed-off-by: Michal Toman <michalt@balena.io> |
| 132 | + author: Michal Toman |
| 133 | + nested: [] |
| 134 | + version: meta-balena-5.3.22 |
| 135 | + title: "" |
| 136 | + date: 2024-07-01T15:01:00.393Z |
| 137 | + - commits: |
| 138 | + - subject: "initrdscripts: Allow passing extra kernel arguments to kexec" |
| 139 | + hash: 02d42d3cc887fb761da348f86bf5e13f1dacaeb2 |
| 140 | + body: > |
| 141 | + At this moment the kexec initrd script just takes the original |
| 142 | + kernel |
| 143 | + |
| 144 | + command line, replaces root with UUID and removes bootloader |
| 145 | + args. |
| 146 | + |
| 147 | + |
| 148 | + We have found at least one use-case (on the Pi4 and firmware |
| 149 | + GPIOs), |
| 150 | + |
| 151 | + where a different initrd script needs to pass extra arguments |
| 152 | + |
| 153 | + to the kexec'd kernel. With this patch it will append the |
| 154 | + contents |
| 155 | + |
| 156 | + of the KEXEC_EXTRA_ARGS variable to the kernel command line. |
| 157 | + footer: |
| 158 | + Change-type: patch |
| 159 | + change-type: patch |
| 160 | + Signed-off-by: Michal Toman <michalt@balena.io> |
| 161 | + signed-off-by: Michal Toman <michalt@balena.io> |
| 162 | + author: Michal Toman |
| 163 | + nested: [] |
| 164 | + version: meta-balena-5.3.21 |
| 165 | + title: "" |
| 166 | + date: 2024-06-13T10:02:03.824Z |
| 167 | + - commits: |
| 168 | + - subject: "hostapp-update-hooks: Re-add check for UEFI to signed-update hook" |
| 169 | + hash: 972ef5f6090fc48a94a313bf01d8d59c6a3570d9 |
| 170 | + body: > |
| 171 | + In 328222014146f0116e0208443f3e255d0e85ef15 we have removed |
| 172 | + |
| 173 | + the signed-update hook from systems that do not have EFI |
| 174 | + |
| 175 | + in MACHINE_FEATURES. This on its own makes sense, however |
| 176 | + together |
| 177 | + |
| 178 | + with it we have also removed the runtime check for whether the |
| 179 | + running |
| 180 | + |
| 181 | + system is actually booted in UEFI mode. |
| 182 | + |
| 183 | + |
| 184 | + This effectively means it is no longer possible to update the |
| 185 | + host OS |
| 186 | + |
| 187 | + on a device type able to boot in both UEFI and BIOS modes |
| 188 | + |
| 189 | + (intel-nuc and genericx86-64-ext) when booted in BIOS mode, |
| 190 | + |
| 191 | + as the signed-update hook is executed unconditionally and fails |
| 192 | + |
| 193 | + if the device is not running UEFI. |
| 194 | + |
| 195 | + |
| 196 | + This patch re-adds the runtime check to only execute the hook |
| 197 | + |
| 198 | + if the system is actually booted in UEFI mode. |
| 199 | + footer: |
| 200 | + Change-type: patch |
| 201 | + change-type: patch |
| 202 | + Signed-off-by: Michal Toman <michalt@balena.io> |
| 203 | + signed-off-by: Michal Toman <michalt@balena.io> |
| 204 | + author: Michal Toman |
| 205 | + nested: [] |
| 206 | + version: meta-balena-5.3.20 |
| 207 | + title: "" |
| 208 | + date: 2024-06-11T12:55:22.727Z |
| 209 | + version: 5.3.23 |
| 210 | + title: "" |
| 211 | + date: 2024-07-03T08:07:29.846Z |
1 | 212 | - commits:
|
2 | 213 | - subject: Update balena-yocto-scripts to da6fe23e89509549866b16cbc6ff404980a189c0
|
3 | 214 | hash: 9a189cdaaadb3f4e325a56c15a9ce66f1b9b0388
|
|
0 commit comments