sn/storage: Fix payload overflow in combined FSTree.Get()

[cthulhu-rider]

i did not try to reproduce this on a cluster. For example, it might show up with

neofs-node/pkg/local_object_storage/writecache/flush.go

Line 224 in 58f146c

err := c.storage.PutBatch(objs)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 25.55%. Comparing base (b3ee421) to head (0d1c686).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3801      +/-   ##
==========================================
+ Coverage   25.52%   25.55%   +0.02%     
==========================================
  Files         661      661              
  Lines       42641    42647       +6     
==========================================
+ Hits        10884    10897      +13     
+ Misses      30753    30748       -5     
+ Partials     1004     1002       -2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[carpawell]

How did you find it?

[roman-khimov]

1. Not sure why this wasn't noticed previously. We have a lot of tests and production networks.
2. IIUC we can have excessive data written into combined file already from write cache (not via regular writes). Which changes the game somewhat, technically the data is still here and your fix as it is now would work fine.
3. Maybe payload-sized LimitReader can be applied layer above.
4. Not sure if we need a tool to fix it for existing fstrees.

[cthulhu-rider]

How did you find it?

Not sure why this wasn't noticed previously. We have a lot of tests and production networks.

noticed by chance while reading this part. Problem was instantly reproduced in unit test, but I didn't try it in devenv. I also didn't see any way the service could mitigate this issue, such as by limiting reading

[roman-khimov]

IIUC we can have excessive data written into combined file already from write cache

Not true, WC doesn't have combined files, so it produces clean buffers that can be written correctly. This means we don't have broken data on disk, which is good.

[cthulhu-rider]

reproduced in devenv

$ neofs-cli object put -r s03.neofs.devenv:8080 -w ../devenv/wallets/wallet.json --cid grnxJmFrMmdEf5ZhzuKaVrdACiDFQbrTeFFxEfYYX5Z --file testfile_16K --no-progress 
Enter password > 
[testfile_16K] Object successfully stored
  OID: 5PMrHej4NCc1J15SPJC6NzRCaHkuwVk6cBEyrwSJVTXr
  CID: grnxJmFrMmdEf5ZhzuKaVrdACiDFQbrTeFFxEfYYX5Z
$
$ neofs-cli object put -r s03.neofs.devenv:8080 -w ../devenv/wallets/wallet.json --cid grnxJmFrMmdEf5ZhzuKaVrdACiDFQbrTeFFxEfYYX5Z --file testfile_16K --no-progress
Enter password > 
[testfile_16K] Object successfully stored
  OID: 4x8nutFJoQ1VFhXbXm5p3H4AA8eTGwQU1a9SMzGv7vZG
  CID: grnxJmFrMmdEf5ZhzuKaVrdACiDFQbrTeFFxEfYYX5Z
$
$ neofs-cli object get -r s03.neofs.devenv:8080 --cid grnxJmFrMmdEf5ZhzuKaVrdACiDFQbrTeFFxEfYYX5Z --oid 4x8nutFJoQ1VFhXbXm5p3H4AA8eTGwQU1a9SMzGv7vZG --ttl 1 --file restored 
 16384 / 16384 [======================================================================================================================================================================================================================================================================================] 100.00% 0s
[restored] Object successfully saved
$
$ neofs-cli object get -r s03.neofs.devenv:8080 --cid grnxJmFrMmdEf5ZhzuKaVrdACiDFQbrTeFFxEfYYX5Z --oid 5PMrHej4NCc1J15SPJC6NzRCaHkuwVk6cBEyrwSJVTXr --ttl 1 --file restored
 32768 / 16384 [======================================================================================================================================================================================================================================================================================] 200.00% 0s
Error: rpc error: copy payload: payload size overflow