The boot banner and the version command print the SDCC release and
revision from the compiler's own macros, and the name of the in-tree
toolchain the Makefile built, which is the checksum of sdcc/ and so
changes with every patch in it. A build with CC given on the command
line says "system" instead, since that compiler need not carry the
patches. A plain release and the same release with the patches report
the same revision, so the toolchain line is what tells them apart.
Costs 66 bytes of BANK2.
Review feedback on #560 (DrDoof):
- A login that came with a cookie that is still valid took a second
slot. The first one looked fresh, so the next login could push out
someone who is using the switch. The slot of the cookie now gets the
new id.
- Changing the password leaves no old login valid: session_init() from
parse_passwd(). With one session the next login pushed a leftover one
out, with several it would stay valid until the next restart.
- The 32-bit comparison of the last use times moves into a leaf function,
so the compiler does not give it a 4-byte scratch in internal RAM.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The web server's bank had no room left for the four sessions once the
original firmware's more robust header handling came in. Picking a slot,
checking a cookie and starting a session move to sessions.c in bank 3;
httpd.c only calls session_init(), session_check() and session_new().
The random id is written straight into the reply, which drops
gen_random_hex_chars() from bank 1.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Logging in from a second browser ended the first one's session at its next
request: two browsers open on the switch logged each other out every few
seconds. Up to four sessions now, each with its own cookie and expiry; a
fifth login replaces the least recently used. The expiry check is a leaf
function so its 32-bit scratch stays out of internal RAM (stack still
159 bytes with HEALTH).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tx2W7xVecMoKbBCkD2i8PG
(cherry picked from commit 03330f4077ef8872d861d951c8e44ece3c573935)
it'll accept just the definition.
Also throughout the regex we have over-specified the number of digits
sometimes ports are 4 digits long, sometimes they are 2. This seems
unnecessary, 1 or more is sufficient
the user clicks apply on the specific vlan. In order for this to
work when the user is simultaneously changing the vlan that their
machine is connected through and the mgmt vlan id, we need to be
able to send a batch of commands through to the switch atomically
so the switch can process all of them and the momentary disconnection
of the machine doesn't matter
When adding a new VLAN, we should load it after addition
Prevent deleting the mgmt vlan
Adding a button to promote a vlan to be the management VLAN
We can't create a vlan called d or mgmt as that clashes with the
commands to delete and promote a vlan
Making the Management VLAN more obvious and adding a "confused"
disabled button to where delete and promote are
Switching to the i image and making it so you can click the entire
row apart from the cell where the buttons are (to prevent accidents)
Cleaning up the last bits of ingress filtering I can see
Also making the first VLAN in the table load when we load the UI
Removing the tickbox for the PVID, I'll swap this out to ticks
or something later on, but the PVID will be automatically set
based on the untagged membership, otherwise it will be 1
Send the vlan with name command when attempting to add a new VLAN
Because of the structure of the vlan command, we need to send the name as well if its set
Set the PVID based on the untagged membership and reload the
table when applied. We need to block setting the PVID to the tagged by proxy
Blue button
Need to provide array of pvid commands not concat
Storing state of each vlan so we can refer later
Blue button
Device page following doc/adding-a-device.md: label data, PCB versus label
naming, CPU identified with 'regget 4' (RTL8373, non-N), separate PHY and flash
identification, build target, PCB silkscreen, port mapping, board signals, front
panel and LED configuration, reset button, MAC location, what works and what is
unverified, plus photos. Row added to supported_devices.md.
RTL8373 (non-N) with an RTL8224 quad 2.5G PHY, 2 MB GD25Q16ESIG flash, no PoE.
The wiring is close to the KP-9000-9XHML-X V3.1 (same 8+1 topology, same I2C on
GPIO39/40), but three things differ and were determined on hardware:
- ModAbs is GPIO30 (low with a module inserted, high with an empty cage) and LOS
is GPIO37 (high without light), not GPIO38. GPIO38, the ModAbs pin of the
KP-9000 reference, is tied low on this board, which made the firmware believe a
module was permanently inserted and retry the I2C read forever.
- The reset button is GPIO54, not GPIO48. GPIO48 never changes here, so the
button appeared to do nothing.
- The LED wiring follows the stock port order (pin 3n+m belongs to port n+1,
LED function m), so the mux table is the chip default except that pin 25 --
the SFP LED -- is driven by function 0, the only populated function in SET1.
Per copper port the left LED shows link/activity and the right LED shows
2.5G, matching the front panel labels.
Verified on hardware: all eight 2.5G ports at 10/100/1000/2500 Mbps, 1G and 10G
SFP modules including DDM data, both LED columns, module insert/remove and the
reset button.
html_min was phony, so every make deleted output/html_min and minified
all of html/ again, even with nothing changed. A stamp file now stands for
the minified copy and depends on the sources and on minify.py. A list of
the source names, rewritten only when it changes, is a dependency too, so
adding or removing a file in html/ still redoes the copy and a removed
file does not stay behind in the image. If output/html_min is missing,
the stamp is remade in the same run.
The note on version.h in webui-compression.md still said it was phony.
Three more bench scenarios around the closing boundary. Every value
of the last image byte, uploaded whole and split right after the
payload, must end the part. A run of CR before the boundary of even
length passed before the fix and an odd one did not, so both are
checked. An image made of CR only, longer than a flash page, and one
made of repeated "\r\n-" exercise the broken matches back to back:
each one appends to the page buffer without a byte written, and the
page flush after both paths has to keep up.
stream_upload() matches the closing multipart boundary byte by byte.
When a byte breaks a partial match, the matched prefix goes to flash as
data, and so does the breaking byte, without being compared against the
start of the boundary.
The boundary starts with "\r\n--", so an image whose last byte is '\r'
never completes: that '\r' starts a match, the '\r' of the boundary
breaks it and is written as data, and the rest of the boundary no
longer matches. The browser waits for a reply that never comes.
Flush the matched prefix and look at the breaking byte again from the
start of the boundary. A boundary may not contain CR, so no longer
suffix of the matched prefix can start a new match.
The 8051 assembler sign-extends every value to 24 bits, so the top
byte of a __sfr32 address list like 0xa4a5a6a7 comes out as 0xff.
The patch keeps 32-bit arithmetic while leaving the 24-bit T line
addresses alone, so the relocations are unchanged. Proposed upstream
in the bug thread. With it SFR_DATA_U32 can be used again, which
saves about 240 bytes of code in REG_SET.
The workflow and the Dockerfile install what building SDCC needs instead
of the distribution's sdcc, and the workflow caches toolchain/ keyed on
the contents of sdcc/.
Released SDCC 4.5.0 and 4.6.0 miscompile some of our code in the mcs51
register allocator (SDCC bugs 4153 and 4154, fixed in trunk r17025 and
r17039). The first make now downloads the 4.6.0 sources, checks their
sha256, applies the two fixes from sdcc/ and installs the compiler into
toolchain/, named after a checksum of sdcc/, so a changed patch builds a
new one. make CC=sdcc still uses an installed compiler.
The masks were 0, so 0x3e2 was ORed into the reset value 0xbff and both
SerDes stayed off on boards without 10G ports (n_10g == 0), e.g. the
HC-SWTG124AS. With SDS 0 off RTL837X_REG_SEC_COUNTER does not hold a
written value: it reads 0, so NTP never retried after a failed first
attempt at boot and froze the clock after a sync.
Tested with regset on an HC-SWTG124AS: the register keeps a write with
SDS 0 in SGMII or 10GR, not with SDS 0 off (also not with only SDS 1
on).
Firefox filled a saved login into the System tab's password fields and
its user name into the NTP poll interval, the input before them. That
input event marked the Time card as edited, so ntpLoad() left NTP off,
the server empty and the time zone at UTC-12:00.
feelfree69
d00f
HL Yi
Brunoz78andClaude Sonnet 5.5
Rob Emery
gtytup
logicog
avicarmeli
robs
Luca S.