Hacktoberfest 2026: as issues que os mantenedores marcaram para outubro, abertas e boas para iniciantes. Ver issues do Hacktoberfest

Agent SIGSEGV on /api/v2|v3/claim — claim_add_user_info_command() calls strchr(NULL) when the lib dir is not writable

Aberta Para iniciantes
  • Algo errado?
#22,786 1 comentário 0 reações 0 responsáveis Ver no GitHub

Mantenedores costumam responder em até 1 dia

Ninguém assumiu esta issue ainda.

Avaliação

Dificuldade
2/5
Tempo estimado
1-3 horas
Facilidade para iniciantes
72/100
Tipo de issue
Bug
Clareza
Claramente especificada
Status de atividade
Pouca atividade
Stack de tecnologia
c, linux
Domínio
api, backend, observability

Direção de pesquisa

Comece em src/web/api/v2/api_v2_claim.c, em claim_add_user_info_command(), depois rastreie netdata_random_session_id_get_filename() e reproduza com a requisição curl fornecida quando o diretório lib não tiver permissão de escrita. Confirme que o endpoint de claim não trava mais quando o nome do arquivo não está disponível e verifique system/systemd/netdata.service.in em busca do gatilho de empacotamento.

Escrita pelo modelo de indexação a partir do texto da issue.

Descrição

Summary

A self-hosted (unclaimed) Netdata Agent crashes with SIGSEGV on every GET /api/v2/claim / GET /api/v3/claim request whenever it cannot create its random-session-id file in netdata_configured_varlib_dir (e.g. when that directory is read-only for the agent process). The web UI issues this request on load / when showing the "Sign-in to continue" panel, so the dashboard reliably crashes the agent in a restart loop.

Root cause is a missing NULL check in claim_add_user_info_command():

// src/web/api/v2/api_v2_claim.c
static void claim_add_user_info_command(BUFFER *wb) {
    const char *filename = netdata_random_session_id_get_filename(); // may be NULL
    ...
    os_filename = filename;          // NULL
    ...
    if(strchr(os_filename, ' '))     // <-- strchr(NULL) => SIGSEGV

netdata_random_session_id_get_filename() returns NULL when netdata_random_session_id_generate() failed (its open(O_WRONLY|O_CREAT...) returned -1 and netdata_random_session_id_filename was left NULL). The caller dereferences it unconditionally.

gdb backtrace (v2.10.3, official Debian .deb)

#0  __strchr_avx2 () at ../sysdeps/x86_64/multiarch/strchr-avx2.S:67
#1  0x... in claim_add_user_info_command (wb=0x...) at src/web/api/v2/api_v2_claim.c:131
#2  claim_json_response (wb=..., response=CLAIM_RESP_INFO, msg=...) at src/web/api/v2/api_v2_claim.c:158
#3  0x... in api_claim (version=..., w=..., url=...) at src/web/api/v2/api_v2_claim.c:230
#4  0x... in web_client_api_request_v3 (...) at src/web/api/web_api_v3.c:269
#5  web_client_api_request (...) at src/web/server/web_client.c:576
#6  check_host_and_call (...) at src/web/server/web_client.c:542
#7  web_client_process_url (... filename = "/api/v3/claim_info" ...) at src/web/server/web_client.c:1163
#8  web_client_process_request_from_web_server (...) 
#9  web_server_rcv_callback (...) at src/web/server/static/static-threaded.c:196
...

bt full for frame #1 shows os_message = "We need to verify this server is yours. SSH to this server and run this command. ...", os_prefix = "sudo cat", and os_filename resolving from the (NULL) session-id filename.

How the file creation fails (the trigger on the official .deb)

netdata_random_session_id_generate() does:

snprintfz(filename, FILENAME_MAX, "%s/netdata_random_session_id", netdata_configured_varlib_dir);
...
int fd = open(filename, O_WRONLY|O_CREAT|O_TRUNC|O_CLOEXEC, 640);
if(fd == -1) { netdata_log_error(...); ret = false; }   // filename stays NULL

On the official Debian package, the shipped systemd unit (system/systemd/netdata.service.in) runs with:

ProtectSystem=full
ProtectHome=read-only
ReadWriteDirectories=/run/netdata
ReadWriteDirectories=-/var/spool/postfix/maildrop

With this sandbox the agent process cannot write to /var/lib/netdata (verified by entering the service's mount namespace):

$ sudo nsenter -t <netdata-pid> -m -- touch /var/lib/netdata/__t
touch: cannot touch '/var/lib/netdata/__t': Read-only file system

So open() returns -1 (EROFS) → netdata_random_session_id_filename stays NULL → get_filename() returns NULL → strchr(NULL) → SIGSEGV. (The same crash would occur on any system where the lib dir is non-writable: full disk, wrong ownership, etc.)

Reproduction

On a v2.10.3 Debian .deb install (agent running as a streaming parent, unclaimed, self-hosted):

# crashes the agent every time:
curl -s -o /dev/null -w '%{http_code}\n' http://127.0.0.1:19999/api/v3/claim
# -> empty reply / 000, journal shows: netdata.service: Main process exited, code=killed, status=11/SEGV

systemctl then restart-loops the agent (Restart=on-failure). Opening the dashboard or clicking "Sign in anonymously" triggers the same request and the same crash.

Suggested fixes

1. Code (the actual crash) — add a NULL guard in claim_add_user_info_command() so the agent never dereferences a NULL session-id filename. (PR incoming.)

2. Packaging (the trigger) — let the agent write its lib dir. Add the lib directory to the unit's writable paths, e.g. in system/systemd/netdata.service.in:

ReadWritePaths=/var/lib/netdata

I verified that adding this drop-in alone makes /api/v3/claim return 200 and the session-id file gets created normally — but the NULL guard is still needed so other "lib dir not writable" conditions can't crash the agent.

Environment

  • Netdata v2.10.3, official .deb (repository.netdata.cloud, stable channel)
  • Debian 12 (bookworm), x86_64, kernel 6.12
  • Agent role: streaming parent for 3 nodes, unclaimed / self-hosted (no Netdata Cloud)
  • netdatacli aclk-state → ACLK Available: Yes

Happy to also send the full bt full / a coredump if useful.

Linguagem predominante
Go
Estrelas
80.8k
Forks
6.6k
Merge médio
17h 49min
PRs com merge (30d)
306

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de netdata/netdata

Todas as issues de netdata/netdata

Issues semelhantes

Mais issues de Go

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.